[{"data":1,"prerenderedAt":1739},["Reactive",2],{"/posts/claude-code-headless":3,"/posts/claude-code-headless:related":1638,"/posts/claude-code-headless:alternate":1738},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"read":11,"body":12,"_type":1632,"_id":1633,"_source":1634,"_file":1635,"_extension":1636,"sitemap":1637},"/posts/claude-code-headless","posts",false,"","Claude Code headless mode: claude -p in scripts and CI","How Claude Code headless mode works: claude -p, text and JSON output, permission flags, bare mode, auth tokens and running it in scripts and GitHub Actions.","2026-11-21","9 min",{"type":13,"children":14,"toc":1613},"root",[15,32,57,64,93,105,110,146,155,194,200,211,213,292,328,337,358,367,387,415,424,459,465,470,498,503,521,530,574,614,630,632,781,800,820,829,835,886,897,906,908,1023,1066,1072,1077,1086,1141,1146,1155,1166,1202,1208,1229,1288,1293,1299,1337,1348,1375,1387,1396,1415,1421,1449,1461,1467,1474,1486,1492,1517,1523,1558,1564,1589,1595],{"type":16,"tag":17,"props":18,"children":19},"element","p",{},[20,23,30],{"type":21,"value":22},"text","Claude Code headless mode is the same agent without the chat window. You run ",{"type":16,"tag":24,"props":25,"children":27},"code",{"className":26},[],[28],{"type":21,"value":29},"claude -p \"prompt\"",{"type":21,"value":31},", it does the work, prints the result and exits with a status code your script can check. This post covers the flags that matter when you use Claude Code headless in shell scripts, cron jobs and CI: output formats, permissions, bare mode, budgets and authentication.",{"type":16,"tag":17,"props":33,"children":34},{},[35,37,46,48,55],{"type":21,"value":36},"All flags here are checked against Anthropic's ",{"type":16,"tag":38,"props":39,"children":43},"a",{"href":40,"rel":41},"https://code.claude.com/docs/en/cli-reference",[42],"noopener",[44],{"type":21,"value":45},"CLI reference",{"type":21,"value":47}," and the ",{"type":16,"tag":38,"props":49,"children":52},{"href":50,"rel":51},"https://code.claude.com/docs/en/headless",[42],[53],{"type":21,"value":54},"headless guide",{"type":21,"value":56},". Some behavior depends on your Claude Code version, and I note the version where the docs do.",{"type":16,"tag":58,"props":59,"children":61},"h2",{"id":60},"what-is-claude-code-headless-mode",[62],{"type":21,"value":63},"What is Claude Code headless mode?",{"type":16,"tag":17,"props":65,"children":66},{},[67,69,75,77,83,85,91],{"type":21,"value":68},"Headless, or non-interactive, mode is what you get when you add ",{"type":16,"tag":24,"props":70,"children":72},{"className":71},[],[73],{"type":21,"value":74},"-p",{"type":21,"value":76}," (long form ",{"type":16,"tag":24,"props":78,"children":80},{"className":79},[],[81],{"type":21,"value":82},"--print",{"type":21,"value":84},") to any ",{"type":16,"tag":24,"props":86,"children":88},{"className":87},[],[89],{"type":21,"value":90},"claude",{"type":21,"value":92}," command. Anthropic now documents it as the CLI form of the Agent SDK: the same tools, agent loop and context handling, available as a command for scripts, or as Python and TypeScript packages when you need more control.",{"type":16,"tag":94,"props":95,"children":100},"pre",{"className":96,"code":98,"language":99,"meta":7},[97],"language-bash","claude -p \"What does the auth module do?\"\n","bash",[101],{"type":16,"tag":24,"props":102,"children":103},{"__ignoreMap":7},[104],{"type":21,"value":98},{"type":16,"tag":17,"props":106,"children":107},{},[108],{"type":21,"value":109},"A few facts that shape how you script it:",{"type":16,"tag":111,"props":112,"children":113},"ul",{},[114,136,141],{"type":16,"tag":115,"props":116,"children":117},"li",{},[118,120,126,128,134],{"type":21,"value":119},"Exit code ",{"type":16,"tag":24,"props":121,"children":123},{"className":122},[],[124],{"type":21,"value":125},"0",{"type":21,"value":127}," means success and non-zero means the run failed, so ",{"type":16,"tag":24,"props":129,"children":131},{"className":130},[],[132],{"type":21,"value":133},"if claude -p ...; then",{"type":21,"value":135}," works.",{"type":16,"tag":115,"props":137,"children":138},{},[139],{"type":21,"value":140},"An invalid flag is reported on stderr before the run starts. A failure inside the run, such as missing authentication, is printed as the result on stdout.",{"type":16,"tag":115,"props":142,"children":143},{},[144],{"type":21,"value":145},"It reads stdin, so you can pipe data in. Piped input is capped at 10 MB. For anything bigger, write it to a file and mention the path in the prompt.",{"type":16,"tag":94,"props":147,"children":150},{"className":148,"code":149,"language":99,"meta":7},[97],"cat build-error.txt | claude -p 'concisely explain the root cause of this build error' > output.txt\n",[151],{"type":16,"tag":24,"props":152,"children":153},{"__ignoreMap":7},[154],{"type":21,"value":149},{"type":16,"tag":17,"props":156,"children":157},{},[158,160,167,169,176,178,184,186,192],{"type":21,"value":159},"On my server, the one I set up with ",{"type":16,"tag":38,"props":161,"children":164},{"href":162,"rel":163},"https://halitsever.github.io/claude-code-vps/",[42],[165],{"type":21,"value":166},"claude-code-vps",{"type":21,"value":168},", I run interactive agents in tmux panes with ",{"type":16,"tag":38,"props":170,"children":173},{"href":171,"rel":172},"https://halitsever.github.io/claudegrid/",[42],[174],{"type":21,"value":175},"claudegrid",{"type":21,"value":177}," when I want to watch and steer them. The full setup is in ",{"type":16,"tag":38,"props":179,"children":181},{"href":180},"/posts/running-claude-code-on-a-server-in-parallel",[182],{"type":21,"value":183},"how I run multiple Claude Code agents in parallel on a server",{"type":21,"value":185},". ",{"type":16,"tag":24,"props":187,"children":189},{"className":188},[],[190],{"type":21,"value":191},"claude -p",{"type":21,"value":193}," covers the other kind of job: one prompt, nobody watching, a result that a script reads.",{"type":16,"tag":58,"props":195,"children":197},{"id":196},"how-do-i-get-json-output-from-claude-code-headless",[198],{"type":21,"value":199},"How do I get JSON output from Claude Code headless?",{"type":16,"tag":17,"props":201,"children":202},{},[203,209],{"type":16,"tag":24,"props":204,"children":206},{"className":205},[],[207],{"type":21,"value":208},"--output-format",{"type":21,"value":210}," takes three values:",{"type":21,"value":212},"\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",{"type":16,"tag":214,"props":215,"children":216},"table",{},[217,237],{"type":16,"tag":218,"props":219,"children":220},"thead",{},[221],{"type":16,"tag":222,"props":223,"children":224},"tr",{},[225,232],{"type":16,"tag":226,"props":227,"children":229},"th",{"align":228},"left",[230],{"type":21,"value":231},"Value",{"type":16,"tag":226,"props":233,"children":234},{"align":228},[235],{"type":21,"value":236},"What you get",{"type":16,"tag":238,"props":239,"children":240},"tbody",{},[241,258,275],{"type":16,"tag":222,"props":242,"children":243},{},[244,253],{"type":16,"tag":245,"props":246,"children":247},"td",{"align":228},[248],{"type":16,"tag":24,"props":249,"children":251},{"className":250},[],[252],{"type":21,"value":21},{"type":16,"tag":245,"props":254,"children":255},{"align":228},[256],{"type":21,"value":257},"Plain text, the default",{"type":16,"tag":222,"props":259,"children":260},{},[261,270],{"type":16,"tag":245,"props":262,"children":263},{"align":228},[264],{"type":16,"tag":24,"props":265,"children":267},{"className":266},[],[268],{"type":21,"value":269},"json",{"type":16,"tag":245,"props":271,"children":272},{"align":228},[273],{"type":21,"value":274},"One JSON object with the result, session ID and metadata",{"type":16,"tag":222,"props":276,"children":277},{},[278,287],{"type":16,"tag":245,"props":279,"children":280},{"align":228},[281],{"type":16,"tag":24,"props":282,"children":284},{"className":283},[],[285],{"type":21,"value":286},"stream-json",{"type":16,"tag":245,"props":288,"children":289},{"align":228},[290],{"type":21,"value":291},"Newline-delimited JSON events as they happen",{"type":16,"tag":17,"props":293,"children":294},{},[295,297,302,304,310,312,318,320,326],{"type":21,"value":296},"With ",{"type":16,"tag":24,"props":298,"children":300},{"className":299},[],[301],{"type":21,"value":269},{"type":21,"value":303},", the text answer is in ",{"type":16,"tag":24,"props":305,"children":307},{"className":306},[],[308],{"type":21,"value":309},"result",{"type":21,"value":311},", and the payload also carries ",{"type":16,"tag":24,"props":313,"children":315},{"className":314},[],[316],{"type":21,"value":317},"session_id",{"type":21,"value":319}," and ",{"type":16,"tag":24,"props":321,"children":323},{"className":322},[],[324],{"type":21,"value":325},"total_cost_usd",{"type":21,"value":327},". The cost is a client-side estimate and can differ from your bill.",{"type":16,"tag":94,"props":329,"children":332},{"className":330,"code":331,"language":99,"meta":7},[97],"claude -p \"Summarize this project\" --output-format json | jq -r '.result'\n",[333],{"type":16,"tag":24,"props":334,"children":335},{"__ignoreMap":7},[336],{"type":21,"value":331},{"type":16,"tag":17,"props":338,"children":339},{},[340,342,348,350,356],{"type":21,"value":341},"When a script needs a fixed shape, add ",{"type":16,"tag":24,"props":343,"children":345},{"className":344},[],[346],{"type":21,"value":347},"--json-schema",{"type":21,"value":349},". The validated output lands in ",{"type":16,"tag":24,"props":351,"children":353},{"className":352},[],[354],{"type":21,"value":355},"structured_output",{"type":21,"value":357},":",{"type":16,"tag":94,"props":359,"children":362},{"className":360,"code":361,"language":99,"meta":7},[97],"claude -p \"Extract the main function names from auth.py\" \\\n  --output-format json \\\n  --json-schema '{\"type\":\"object\",\"properties\":{\"functions\":{\"type\":\"array\",\"items\":{\"type\":\"string\"}}},\"required\":[\"functions\"]}' \\\n  | jq '.structured_output'\n",[363],{"type":16,"tag":24,"props":364,"children":365},{"__ignoreMap":7},[366],{"type":21,"value":361},{"type":16,"tag":17,"props":368,"children":369},{},[370,372,377,379,385],{"type":21,"value":371},"If Claude declines or the run ends without valid structured output, the field is missing. Check for it before you use it. An invalid schema makes ",{"type":16,"tag":24,"props":373,"children":375},{"className":374},[],[376],{"type":21,"value":90},{"type":21,"value":378}," exit with ",{"type":16,"tag":24,"props":380,"children":382},{"className":381},[],[383],{"type":21,"value":384},"Error: --json-schema is not a valid JSON Schema",{"type":21,"value":386},".",{"type":16,"tag":17,"props":388,"children":389},{},[390,392,397,399,405,407,413],{"type":21,"value":391},"For live output, the docs pair ",{"type":16,"tag":24,"props":393,"children":395},{"className":394},[],[396],{"type":21,"value":286},{"type":21,"value":398}," with ",{"type":16,"tag":24,"props":400,"children":402},{"className":401},[],[403],{"type":21,"value":404},"--verbose",{"type":21,"value":406},", and ",{"type":16,"tag":24,"props":408,"children":410},{"className":409},[],[411],{"type":21,"value":412},"--include-partial-messages",{"type":21,"value":414}," adds token-level events:",{"type":16,"tag":94,"props":416,"children":419},{"className":417,"code":418,"language":99,"meta":7},[97],"claude -p \"Explain recursion\" --output-format stream-json --verbose --include-partial-messages | \\\n  jq -rj 'select(.type == \"stream_event\" and .event.delta.type? == \"text_delta\") | .event.delta.text'\n",[420],{"type":16,"tag":24,"props":421,"children":422},{"__ignoreMap":7},[423],{"type":21,"value":418},{"type":16,"tag":17,"props":425,"children":426},{},[427,429,434,436,442,444,450,451,457],{"type":21,"value":428},"The last line of the stream is a ",{"type":16,"tag":24,"props":430,"children":432},{"className":431},[],[433],{"type":21,"value":309},{"type":21,"value":435}," message with the final text, cost and session metadata. The first is usually a ",{"type":16,"tag":24,"props":437,"children":439},{"className":438},[],[440],{"type":21,"value":441},"system/init",{"type":21,"value":443}," event listing the model, tools, MCP servers and plugins, which is handy in CI: its ",{"type":16,"tag":24,"props":445,"children":447},{"className":446},[],[448],{"type":21,"value":449},"plugin_errors",{"type":21,"value":319},{"type":16,"tag":24,"props":452,"children":454},{"className":453},[],[455],{"type":21,"value":456},"mcp_server_errors",{"type":21,"value":458}," arrays let a job fail when something didn't load.",{"type":16,"tag":58,"props":460,"children":462},{"id":461},"permissions-in-headless-mode",[463],{"type":21,"value":464},"Permissions in headless mode",{"type":16,"tag":17,"props":466,"children":467},{},[468],{"type":21,"value":469},"In headless mode nobody is there to answer a permission prompt, so what matters is what happens to a tool call that would normally ask.",{"type":16,"tag":17,"props":471,"children":472},{},[473,475,480,482,488,490,496],{"type":21,"value":474},"In Claude Code v2.1.302 or later, a ",{"type":16,"tag":24,"props":476,"children":478},{"className":477},[],[479],{"type":21,"value":191},{"type":21,"value":481}," run with no permission mode set starts in Manual mode (",{"type":16,"tag":24,"props":483,"children":485},{"className":484},[],[486],{"type":21,"value":487},"default",{"type":21,"value":489}," in config). A call that needs approval is denied unless an allow rule, a hook or a ",{"type":16,"tag":24,"props":491,"children":493},{"className":492},[],[494],{"type":21,"value":495},"--permission-prompt-tool",{"type":21,"value":497}," approves it. Reads in your working directories still run, since they need no approval.",{"type":16,"tag":17,"props":499,"children":500},{},[501],{"type":21,"value":502},"You have three levels of control.",{"type":16,"tag":17,"props":504,"children":505},{},[506,512,513,519],{"type":16,"tag":507,"props":508,"children":509},"strong",{},[510],{"type":21,"value":511},"Allow specific tools",{"type":21,"value":398},{"type":16,"tag":24,"props":514,"children":516},{"className":515},[],[517],{"type":21,"value":518},"--allowedTools",{"type":21,"value":520},". It uses permission rule syntax, so you can scope Bash to commands:",{"type":16,"tag":94,"props":522,"children":525},{"className":523,"code":524,"language":99,"meta":7},[97],"claude -p \"Look at my staged changes and create an appropriate commit\" \\\n  --allowedTools \"Bash(git diff *),Bash(git log *),Bash(git status *),Bash(git commit *)\"\n",[526],{"type":16,"tag":24,"props":527,"children":528},{"__ignoreMap":7},[529],{"type":21,"value":524},{"type":16,"tag":17,"props":531,"children":532},{},[533,535,541,543,549,551,557,559,565,567,573],{"type":21,"value":534},"The space before ",{"type":16,"tag":24,"props":536,"children":538},{"className":537},[],[539],{"type":21,"value":540},"*",{"type":21,"value":542}," matters. ",{"type":16,"tag":24,"props":544,"children":546},{"className":545},[],[547],{"type":21,"value":548},"Bash(git diff *)",{"type":21,"value":550}," matches ",{"type":16,"tag":24,"props":552,"children":554},{"className":553},[],[555],{"type":21,"value":556},"git diff",{"type":21,"value":558}," with or without arguments, while ",{"type":16,"tag":24,"props":560,"children":562},{"className":561},[],[563],{"type":21,"value":564},"Bash(git diff*)",{"type":21,"value":566}," would also match ",{"type":16,"tag":24,"props":568,"children":570},{"className":569},[],[571],{"type":21,"value":572},"git diff-index",{"type":21,"value":386},{"type":16,"tag":17,"props":575,"children":576},{},[577,582,583,589,591,597,599,605,607,613],{"type":16,"tag":507,"props":578,"children":579},{},[580],{"type":21,"value":581},"Block or remove tools",{"type":21,"value":398},{"type":16,"tag":24,"props":584,"children":586},{"className":585},[],[587],{"type":21,"value":588},"--disallowedTools",{"type":21,"value":590}," (deny rules, for example ",{"type":16,"tag":24,"props":592,"children":594},{"className":593},[],[595],{"type":21,"value":596},"\"Bash(rm *)\"",{"type":21,"value":598}," or ",{"type":16,"tag":24,"props":600,"children":602},{"className":601},[],[603],{"type":21,"value":604},"\"mcp__*\"",{"type":21,"value":606},") or restrict the built-in set with ",{"type":16,"tag":24,"props":608,"children":610},{"className":609},[],[611],{"type":21,"value":612},"--tools \"Read,Grep\"",{"type":21,"value":386},{"type":16,"tag":17,"props":615,"children":616},{},[617,622,623,629],{"type":16,"tag":507,"props":618,"children":619},{},[620],{"type":21,"value":621},"Set a baseline",{"type":21,"value":398},{"type":16,"tag":24,"props":624,"children":626},{"className":625},[],[627],{"type":21,"value":628},"--permission-mode",{"type":21,"value":357},{"type":21,"value":631},"\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",{"type":16,"tag":214,"props":633,"children":634},{},[635,656],{"type":16,"tag":218,"props":636,"children":637},{},[638],{"type":16,"tag":222,"props":639,"children":640},{},[641,646,651],{"type":16,"tag":226,"props":642,"children":643},{"align":228},[644],{"type":21,"value":645},"Mode",{"type":16,"tag":226,"props":647,"children":648},{"align":228},[649],{"type":21,"value":650},"What runs without asking",{"type":16,"tag":226,"props":652,"children":653},{"align":228},[654],{"type":21,"value":655},"Fits",{"type":16,"tag":238,"props":657,"children":658},{},[659,680,715,737,759],{"type":16,"tag":222,"props":660,"children":661},{},[662,670,675],{"type":16,"tag":245,"props":663,"children":664},{"align":228},[665],{"type":16,"tag":24,"props":666,"children":668},{"className":667},[],[669],{"type":21,"value":487},{"type":16,"tag":245,"props":671,"children":672},{"align":228},[673],{"type":21,"value":674},"Reads and actions that need no approval",{"type":16,"tag":245,"props":676,"children":677},{"align":228},[678],{"type":21,"value":679},"Read-only reports",{"type":16,"tag":222,"props":681,"children":682},{},[683,692,710],{"type":16,"tag":245,"props":684,"children":685},{"align":228},[686],{"type":16,"tag":24,"props":687,"children":689},{"className":688},[],[690],{"type":21,"value":691},"acceptEdits",{"type":16,"tag":245,"props":693,"children":694},{"align":228},[695,697,703,704],{"type":21,"value":696},"File edits and common filesystem commands like ",{"type":16,"tag":24,"props":698,"children":700},{"className":699},[],[701],{"type":21,"value":702},"mkdir",{"type":21,"value":319},{"type":16,"tag":24,"props":705,"children":707},{"className":706},[],[708],{"type":21,"value":709},"mv",{"type":16,"tag":245,"props":711,"children":712},{"align":228},[713],{"type":21,"value":714},"Applying fixes in a checkout",{"type":16,"tag":222,"props":716,"children":717},{},[718,727,732],{"type":16,"tag":245,"props":719,"children":720},{"align":228},[721],{"type":16,"tag":24,"props":722,"children":724},{"className":723},[],[725],{"type":21,"value":726},"dontAsk",{"type":16,"tag":245,"props":728,"children":729},{"align":228},[730],{"type":21,"value":731},"Only pre-approved tools. Anything that would prompt is denied",{"type":16,"tag":245,"props":733,"children":734},{"align":228},[735],{"type":21,"value":736},"Locked-down CI",{"type":16,"tag":222,"props":738,"children":739},{},[740,749,754],{"type":16,"tag":245,"props":741,"children":742},{"align":228},[743],{"type":16,"tag":24,"props":744,"children":746},{"className":745},[],[747],{"type":21,"value":748},"auto",{"type":16,"tag":245,"props":750,"children":751},{"align":228},[752],{"type":21,"value":753},"Most actions, reviewed by a separate classifier model",{"type":16,"tag":245,"props":755,"children":756},{"align":228},[757],{"type":21,"value":758},"Longer unattended tasks",{"type":16,"tag":222,"props":760,"children":761},{},[762,771,776],{"type":16,"tag":245,"props":763,"children":764},{"align":228},[765],{"type":16,"tag":24,"props":766,"children":768},{"className":767},[],[769],{"type":21,"value":770},"bypassPermissions",{"type":16,"tag":245,"props":772,"children":773},{"align":228},[774],{"type":21,"value":775},"Everything",{"type":16,"tag":245,"props":777,"children":778},{"align":228},[779],{"type":21,"value":780},"Isolated containers and VMs only",{"type":16,"tag":17,"props":782,"children":783},{},[784,790,792,798],{"type":16,"tag":24,"props":785,"children":787},{"className":786},[],[788],{"type":21,"value":789},"--dangerously-skip-permissions",{"type":21,"value":791}," is the same as ",{"type":16,"tag":24,"props":793,"children":795},{"className":794},[],[796],{"type":21,"value":797},"--permission-mode bypassPermissions",{"type":21,"value":799},". The docs say to use it only in isolated environments like containers, VMs or dev containers without internet access, where Claude Code can't damage the host. I'd take that literally.",{"type":16,"tag":17,"props":801,"children":802},{},[803,805,811,813,819],{"type":21,"value":804},"For scheduled jobs there is also ",{"type":16,"tag":24,"props":806,"children":808},{"className":807},[],[809],{"type":21,"value":810},"--permission-prompts none",{"type":21,"value":812}," (v2.1.259 or later). It tells Claude that nobody can approve a request, so it shouldn't retry, and it removes tools that need a person, such as ",{"type":16,"tag":24,"props":814,"children":816},{"className":815},[],[817],{"type":21,"value":818},"AskUserQuestion",{"type":21,"value":357},{"type":16,"tag":94,"props":821,"children":824},{"className":822,"code":823,"language":99,"meta":7},[97],"claude -p \"Update the dependency pins and run the tests\" --permission-mode auto --permission-prompts none\n",[825],{"type":16,"tag":24,"props":826,"children":827},{"__ignoreMap":7},[828],{"type":21,"value":823},{"type":16,"tag":58,"props":830,"children":832},{"id":831},"bare-mode-the-safer-default-for-scripts",[833],{"type":21,"value":834},"Bare mode: the safer default for scripts",{"type":16,"tag":17,"props":836,"children":837},{},[838,840,845,847,853,855,861,863,869,871,877,879,884],{"type":21,"value":839},"Without extra flags, ",{"type":16,"tag":24,"props":841,"children":843},{"className":842},[],[844],{"type":21,"value":191},{"type":21,"value":846}," loads everything an interactive session would: your hooks, skills, plugins, MCP servers, auto memory and ",{"type":16,"tag":24,"props":848,"children":850},{"className":849},[],[851],{"type":21,"value":852},"CLAUDE.md",{"type":21,"value":854},", from both the project and ",{"type":16,"tag":24,"props":856,"children":858},{"className":857},[],[859],{"type":21,"value":860},"~/.claude",{"type":21,"value":862},". That includes a project's ",{"type":16,"tag":24,"props":864,"children":866},{"className":865},[],[867],{"type":21,"value":868},".claude/settings.json",{"type":21,"value":870}," hooks and ",{"type":16,"tag":24,"props":872,"children":874},{"className":873},[],[875],{"type":21,"value":876},".mcp.json",{"type":21,"value":878}," servers, even in a folder you've never trusted, and a ",{"type":16,"tag":24,"props":880,"children":882},{"className":881},[],[883],{"type":21,"value":74},{"type":21,"value":885}," run shows no trust dialog.",{"type":16,"tag":17,"props":887,"children":888},{},[889,895],{"type":16,"tag":24,"props":890,"children":892},{"className":891},[],[893],{"type":21,"value":894},"--bare",{"type":21,"value":896}," skips all of that auto-discovery and starts faster. Claude gets Bash, file read and file edit tools, and you pass in exactly what the run needs:",{"type":16,"tag":94,"props":898,"children":901},{"className":899,"code":900,"language":99,"meta":7},[97],"claude --bare -p \"Summarize README.md\" --allowedTools \"Read\"\n",[902],{"type":16,"tag":24,"props":903,"children":904},{"__ignoreMap":7},[905],{"type":21,"value":900},{"type":21,"value":907},"\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",{"type":16,"tag":214,"props":909,"children":910},{},[911,927],{"type":16,"tag":218,"props":912,"children":913},{},[914],{"type":16,"tag":222,"props":915,"children":916},{},[917,922],{"type":16,"tag":226,"props":918,"children":919},{"align":228},[920],{"type":21,"value":921},"To load",{"type":16,"tag":226,"props":923,"children":924},{"align":228},[925],{"type":21,"value":926},"Use",{"type":16,"tag":238,"props":928,"children":929},{},[930,955,972,989,1006],{"type":16,"tag":222,"props":931,"children":932},{},[933,938],{"type":16,"tag":245,"props":934,"children":935},{"align":228},[936],{"type":21,"value":937},"Extra system prompt",{"type":16,"tag":245,"props":939,"children":940},{"align":228},[941,947,949],{"type":16,"tag":24,"props":942,"children":944},{"className":943},[],[945],{"type":21,"value":946},"--append-system-prompt",{"type":21,"value":948},", ",{"type":16,"tag":24,"props":950,"children":952},{"className":951},[],[953],{"type":21,"value":954},"--append-system-prompt-file",{"type":16,"tag":222,"props":956,"children":957},{},[958,963],{"type":16,"tag":245,"props":959,"children":960},{"align":228},[961],{"type":21,"value":962},"Settings",{"type":16,"tag":245,"props":964,"children":965},{"align":228},[966],{"type":16,"tag":24,"props":967,"children":969},{"className":968},[],[970],{"type":21,"value":971},"--settings \u003Cfile-or-json>",{"type":16,"tag":222,"props":973,"children":974},{},[975,980],{"type":16,"tag":245,"props":976,"children":977},{"align":228},[978],{"type":21,"value":979},"MCP servers",{"type":16,"tag":245,"props":981,"children":982},{"align":228},[983],{"type":16,"tag":24,"props":984,"children":986},{"className":985},[],[987],{"type":21,"value":988},"--mcp-config \u003Cfile-or-json>",{"type":16,"tag":222,"props":990,"children":991},{},[992,997],{"type":16,"tag":245,"props":993,"children":994},{"align":228},[995],{"type":21,"value":996},"Subagents",{"type":16,"tag":245,"props":998,"children":999},{"align":228},[1000],{"type":16,"tag":24,"props":1001,"children":1003},{"className":1002},[],[1004],{"type":21,"value":1005},"--agents \u003Cfile-or-json>",{"type":16,"tag":222,"props":1007,"children":1008},{},[1009,1014],{"type":16,"tag":245,"props":1010,"children":1011},{"align":228},[1012],{"type":21,"value":1013},"A plugin",{"type":16,"tag":245,"props":1015,"children":1016},{"align":228},[1017],{"type":16,"tag":24,"props":1018,"children":1020},{"className":1019},[],[1021],{"type":21,"value":1022},"--plugin-dir \u003Cpath>",{"type":16,"tag":17,"props":1024,"children":1025},{},[1026,1028,1033,1035,1040,1042,1048,1050,1056,1058,1064],{"type":21,"value":1027},"Anthropic calls ",{"type":16,"tag":24,"props":1029,"children":1031},{"className":1030},[],[1032],{"type":21,"value":894},{"type":21,"value":1034}," the recommended mode for scripted and SDK calls and says it will become the default for ",{"type":16,"tag":24,"props":1036,"children":1038},{"className":1037},[],[1039],{"type":21,"value":74},{"type":21,"value":1041}," in a future release. The catch is authentication: bare mode never reads OAuth credentials or the keychain, so it needs ",{"type":16,"tag":24,"props":1043,"children":1045},{"className":1044},[],[1046],{"type":21,"value":1047},"ANTHROPIC_API_KEY",{"type":21,"value":1049}," or an ",{"type":16,"tag":24,"props":1051,"children":1053},{"className":1052},[],[1054],{"type":21,"value":1055},"apiKeyHelper",{"type":21,"value":1057}," in the ",{"type":16,"tag":24,"props":1059,"children":1061},{"className":1060},[],[1062],{"type":21,"value":1063},"--settings",{"type":21,"value":1065}," JSON.",{"type":16,"tag":58,"props":1067,"children":1069},{"id":1068},"keeping-a-headless-run-bounded",[1070],{"type":21,"value":1071},"Keeping a headless run bounded",{"type":16,"tag":17,"props":1073,"children":1074},{},[1075],{"type":21,"value":1076},"A script shouldn't be able to run forever or spend without limit. These flags are print-mode only:",{"type":16,"tag":94,"props":1078,"children":1081},{"className":1079,"code":1080,"language":99,"meta":7},[97],"claude -p \"Fix the failing test in tests/auth.test.ts\" \\\n  --allowedTools \"Read,Edit,Bash(npm test *)\" \\\n  --max-turns 10 \\\n  --max-budget-usd 2 \\\n  --no-session-persistence\n",[1082],{"type":16,"tag":24,"props":1083,"children":1084},{"__ignoreMap":7},[1085],{"type":21,"value":1080},{"type":16,"tag":111,"props":1087,"children":1088},{},[1089,1100,1111,1122],{"type":16,"tag":115,"props":1090,"children":1091},{},[1092,1098],{"type":16,"tag":24,"props":1093,"children":1095},{"className":1094},[],[1096],{"type":21,"value":1097},"--max-turns",{"type":21,"value":1099}," limits agentic turns. There is no limit by default.",{"type":16,"tag":115,"props":1101,"children":1102},{},[1103,1109],{"type":16,"tag":24,"props":1104,"children":1106},{"className":1105},[],[1107],{"type":21,"value":1108},"--max-budget-usd",{"type":21,"value":1110}," stops the run once estimated spend reaches the amount. Spend can pass the cap, so leave headroom.",{"type":16,"tag":115,"props":1112,"children":1113},{},[1114,1120],{"type":16,"tag":24,"props":1115,"children":1117},{"className":1116},[],[1118],{"type":21,"value":1119},"--no-session-persistence",{"type":21,"value":1121}," keeps the session off disk, so it can't be resumed.",{"type":16,"tag":115,"props":1123,"children":1124},{},[1125,1131,1133,1139],{"type":16,"tag":24,"props":1126,"children":1128},{"className":1127},[],[1129],{"type":21,"value":1130},"--model",{"type":21,"value":1132}," picks the model, and ",{"type":16,"tag":24,"props":1134,"children":1136},{"className":1135},[],[1137],{"type":21,"value":1138},"--fallback-model sonnet,haiku",{"type":21,"value":1140}," gives a chain to try when the primary is overloaded or unavailable.",{"type":16,"tag":17,"props":1142,"children":1143},{},[1144],{"type":21,"value":1145},"To continue work across runs, save the session ID and resume it:",{"type":16,"tag":94,"props":1147,"children":1150},{"className":1148,"code":1149,"language":99,"meta":7},[97],"session_id=$(claude -p \"Start a review of src/\" --output-format json | jq -r '.session_id')\nclaude -p \"Now focus on the database queries\" --resume \"$session_id\"\n",[1151],{"type":16,"tag":24,"props":1152,"children":1153},{"__ignoreMap":7},[1154],{"type":21,"value":1149},{"type":16,"tag":17,"props":1156,"children":1157},{},[1158,1164],{"type":16,"tag":24,"props":1159,"children":1161},{"className":1160},[],[1162],{"type":21,"value":1163},"--continue",{"type":21,"value":1165}," picks up the most recent conversation instead.",{"type":16,"tag":17,"props":1167,"children":1168},{},[1169,1171,1177,1179,1184,1186,1192,1194,1200],{"type":21,"value":1170},"Skills work too. Put ",{"type":16,"tag":24,"props":1172,"children":1174},{"className":1173},[],[1175],{"type":21,"value":1176},"/skill-name",{"type":21,"value":1178}," in the prompt string and Claude Code expands it before running. If you stack several, only the first loads in ",{"type":16,"tag":24,"props":1180,"children":1182},{"className":1181},[],[1183],{"type":21,"value":74},{"type":21,"value":1185},", with the rest of the line as its arguments. A skill that runs in a subagent with ",{"type":16,"tag":24,"props":1187,"children":1189},{"className":1188},[],[1190],{"type":21,"value":1191},"context: fork",{"type":21,"value":1193}," is waited for instead of running in the background. My post on ",{"type":16,"tag":38,"props":1195,"children":1197},{"href":1196},"/posts/claude-code-skills",[1198],{"type":21,"value":1199},"Claude Code skills",{"type":21,"value":1201}," covers how to write one.",{"type":16,"tag":58,"props":1203,"children":1205},{"id":1204},"authentication-on-a-server-or-in-ci",[1206],{"type":21,"value":1207},"Authentication on a server or in CI",{"type":16,"tag":17,"props":1209,"children":1210},{},[1211,1213,1219,1221,1228],{"type":21,"value":1212},"There is no browser on a CI runner, so ",{"type":16,"tag":24,"props":1214,"children":1216},{"className":1215},[],[1217],{"type":21,"value":1218},"/login",{"type":21,"value":1220}," is out. Two options from the ",{"type":16,"tag":38,"props":1222,"children":1225},{"href":1223,"rel":1224},"https://code.claude.com/docs/en/authentication",[42],[1226],{"type":21,"value":1227},"authentication docs",{"type":21,"value":357},{"type":16,"tag":111,"props":1230,"children":1231},{},[1232,1255],{"type":16,"tag":115,"props":1233,"children":1234},{},[1235,1240,1242,1247,1249,1254],{"type":16,"tag":507,"props":1236,"children":1237},{},[1238],{"type":21,"value":1239},"API key",{"type":21,"value":1241},": set ",{"type":16,"tag":24,"props":1243,"children":1245},{"className":1244},[],[1246],{"type":21,"value":1047},{"type":21,"value":1248}," with a key from the Claude Console. This is the one that works with ",{"type":16,"tag":24,"props":1250,"children":1252},{"className":1251},[],[1253],{"type":21,"value":894},{"type":21,"value":386},{"type":16,"tag":115,"props":1256,"children":1257},{},[1258,1263,1265,1271,1273,1278,1280,1286],{"type":16,"tag":507,"props":1259,"children":1260},{},[1261],{"type":21,"value":1262},"Subscription token",{"type":21,"value":1264},": run ",{"type":16,"tag":24,"props":1266,"children":1268},{"className":1267},[],[1269],{"type":21,"value":1270},"claude setup-token",{"type":21,"value":1272}," once on a machine with a browser. It requires a Claude subscription, goes through the same approval as ",{"type":16,"tag":24,"props":1274,"children":1276},{"className":1275},[],[1277],{"type":21,"value":1218},{"type":21,"value":1279}," and prints a one-year OAuth token without saving it anywhere. Set it as ",{"type":16,"tag":24,"props":1281,"children":1283},{"className":1282},[],[1284],{"type":21,"value":1285},"CLAUDE_CODE_OAUTH_TOKEN",{"type":21,"value":1287}," where the job runs. Bare mode doesn't read this variable.",{"type":16,"tag":17,"props":1289,"children":1290},{},[1291],{"type":21,"value":1292},"Either way, keep the value in your CI's secret store and never in the repository.",{"type":16,"tag":58,"props":1294,"children":1296},{"id":1295},"claude-code-headless-in-github-actions",[1297],{"type":21,"value":1298},"Claude Code headless in GitHub Actions",{"type":16,"tag":17,"props":1300,"children":1301},{},[1302,1304,1311,1313,1319,1321,1327,1329,1335],{"type":21,"value":1303},"The simplest route is Anthropic's ",{"type":16,"tag":38,"props":1305,"children":1308},{"href":1306,"rel":1307},"https://code.claude.com/docs/en/github-actions",[42],[1309],{"type":21,"value":1310},"Claude Code GitHub Action",{"type":21,"value":1312},". Running ",{"type":16,"tag":24,"props":1314,"children":1316},{"className":1315},[],[1317],{"type":21,"value":1318},"/install-github-app",{"type":21,"value":1320}," inside Claude Code sets up the GitHub app, the secret and a workflow for you, on github.com repositories. With a ",{"type":16,"tag":24,"props":1322,"children":1324},{"className":1323},[],[1325],{"type":21,"value":1326},"prompt",{"type":21,"value":1328}," input the action runs in automation mode on any event, including a schedule, and ",{"type":16,"tag":24,"props":1330,"children":1332},{"className":1331},[],[1333],{"type":21,"value":1334},"claude_args",{"type":21,"value":1336}," passes CLI flags through:",{"type":16,"tag":94,"props":1338,"children":1343},{"className":1339,"code":1341,"language":1342,"meta":7},[1340],"language-yaml","name: Daily Report\non:\n  schedule:\n    - cron: \"0 9 * * *\"\njobs:\n  report:\n    runs-on: ubuntu-latest\n    permissions:\n      contents: read\n      issues: read\n      id-token: write\n    steps:\n      - uses: anthropics/claude-code-action@v1\n        with:\n          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}\n          prompt: \"Generate a summary of yesterday's commits and open issues\"\n          claude_args: |\n            --allowedTools \"mcp__github__list_commits,mcp__github__list_issues\"\n","yaml",[1344],{"type":16,"tag":24,"props":1345,"children":1346},{"__ignoreMap":7},[1347],{"type":21,"value":1341},{"type":16,"tag":17,"props":1349,"children":1350},{},[1351,1353,1358,1360,1366,1368,1374],{"type":21,"value":1352},"With a plain-text prompt, Claude has no shell or GitHub API access until you grant the tools it needs, as ",{"type":16,"tag":24,"props":1354,"children":1356},{"className":1355},[],[1357],{"type":21,"value":518},{"type":21,"value":1359}," does here. For a subscription token, swap the ",{"type":16,"tag":24,"props":1361,"children":1363},{"className":1362},[],[1364],{"type":21,"value":1365},"anthropic_api_key",{"type":21,"value":1367}," line for ",{"type":16,"tag":24,"props":1369,"children":1371},{"className":1370},[],[1372],{"type":21,"value":1373},"claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}",{"type":21,"value":386},{"type":16,"tag":17,"props":1376,"children":1377},{},[1378,1380,1385],{"type":21,"value":1379},"If you'd rather call the CLI yourself on a runner or your own server, install it with the native installer and run ",{"type":16,"tag":24,"props":1381,"children":1383},{"className":1382},[],[1384],{"type":21,"value":191},{"type":21,"value":1386}," like any other command:",{"type":16,"tag":94,"props":1388,"children":1391},{"className":1389,"code":1390,"language":99,"meta":7},[97],"curl -fsSL https://claude.ai/install.sh | bash\ngit diff origin/main | claude --bare -p \"Review this diff for bugs. Reply with file:line and the issue, nothing else.\" \\\n  --output-format json --max-turns 5 | jq -r '.result'\n",[1392],{"type":16,"tag":24,"props":1393,"children":1394},{"__ignoreMap":7},[1395],{"type":21,"value":1390},{"type":16,"tag":17,"props":1397,"children":1398},{},[1399,1401,1406,1408,1413],{"type":21,"value":1400},"This one uses ",{"type":16,"tag":24,"props":1402,"children":1404},{"className":1403},[],[1405],{"type":21,"value":894},{"type":21,"value":1407},", so the job needs ",{"type":16,"tag":24,"props":1409,"children":1411},{"className":1410},[],[1412],{"type":21,"value":1047},{"type":21,"value":1414}," in its environment. Piping the diff in means Claude doesn't need Bash permission to read it.",{"type":16,"tag":58,"props":1416,"children":1418},{"id":1417},"stopping-a-run-cleanly",[1419],{"type":21,"value":1420},"Stopping a run cleanly",{"type":16,"tag":17,"props":1422,"children":1423},{},[1424,1426,1431,1433,1439,1441,1447],{"type":21,"value":1425},"If a supervisor stops ",{"type":16,"tag":24,"props":1427,"children":1429},{"className":1428},[],[1430],{"type":21,"value":191},{"type":21,"value":1432}," with SIGTERM, it exits with code 143, kills the process tree of any running Bash command, runs ",{"type":16,"tag":24,"props":1434,"children":1436},{"className":1435},[],[1437],{"type":21,"value":1438},"SessionEnd",{"type":21,"value":1440}," hooks and leaves the turn unfinished with no result. Send SIGINT first if you want the turn to end properly. After the turn, a run may stay open to wait for background work Claude started. By default Claude Code stops waiting after 10 minutes of idle time, and ",{"type":16,"tag":24,"props":1442,"children":1444},{"className":1443},[],[1445],{"type":21,"value":1446},"CLAUDE_CODE_PRINT_BG_WAIT_CEILING_MS",{"type":21,"value":1448}," changes that cap.",{"type":16,"tag":17,"props":1450,"children":1451},{},[1452,1454,1460],{"type":21,"value":1453},"If you need this wired into a product or pipeline, I take on that kind of work. See ",{"type":16,"tag":38,"props":1455,"children":1457},{"href":1456},"/hire",[1458],{"type":21,"value":1459},"hire",{"type":21,"value":386},{"type":16,"tag":58,"props":1462,"children":1464},{"id":1463},"faq",[1465],{"type":21,"value":1466},"FAQ",{"type":16,"tag":1468,"props":1469,"children":1471},"h3",{"id":1470},"is-claude-code-headless-mode-the-same-as-the-agent-sdk",[1472],{"type":21,"value":1473},"Is Claude Code headless mode the same as the Agent SDK?",{"type":16,"tag":17,"props":1475,"children":1476},{},[1477,1479,1484],{"type":21,"value":1478},"Anthropic documents ",{"type":16,"tag":24,"props":1480,"children":1482},{"className":1481},[],[1483],{"type":21,"value":191},{"type":21,"value":1485}," as the Agent SDK used through the CLI. The Python and TypeScript packages add things like tool approval callbacks and native message objects.",{"type":16,"tag":1468,"props":1487,"children":1489},{"id":1488},"why-are-my-tool-calls-denied-in-claude-p",[1490],{"type":21,"value":1491},"Why are my tool calls denied in claude -p?",{"type":16,"tag":17,"props":1493,"children":1494},{},[1495,1497,1502,1504,1509,1511,1516],{"type":21,"value":1496},"Recent versions start ",{"type":16,"tag":24,"props":1498,"children":1500},{"className":1499},[],[1501],{"type":21,"value":74},{"type":21,"value":1503}," runs in Manual mode, where anything that needs approval is denied. Add ",{"type":16,"tag":24,"props":1505,"children":1507},{"className":1506},[],[1508],{"type":21,"value":518},{"type":21,"value":1510}," entries or pick a mode with ",{"type":16,"tag":24,"props":1512,"children":1514},{"className":1513},[],[1515],{"type":21,"value":628},{"type":21,"value":386},{"type":16,"tag":1468,"props":1518,"children":1520},{"id":1519},"how-do-i-get-only-the-answer-text-from-a-headless-run",[1521],{"type":21,"value":1522},"How do I get only the answer text from a headless run?",{"type":16,"tag":17,"props":1524,"children":1525},{},[1526,1528,1534,1536,1541,1543,1549,1551,1556],{"type":21,"value":1527},"Use ",{"type":16,"tag":24,"props":1529,"children":1531},{"className":1530},[],[1532],{"type":21,"value":1533},"--output-format json",{"type":21,"value":1535}," and read the ",{"type":16,"tag":24,"props":1537,"children":1539},{"className":1538},[],[1540],{"type":21,"value":309},{"type":21,"value":1542}," field with ",{"type":16,"tag":24,"props":1544,"children":1546},{"className":1545},[],[1547],{"type":21,"value":1548},"jq -r '.result'",{"type":21,"value":1550},", or keep the default ",{"type":16,"tag":24,"props":1552,"children":1554},{"className":1553},[],[1555],{"type":21,"value":21},{"type":21,"value":1557}," format.",{"type":16,"tag":1468,"props":1559,"children":1561},{"id":1560},"can-i-use-my-claude-subscription-in-ci",[1562],{"type":21,"value":1563},"Can I use my Claude subscription in CI?",{"type":16,"tag":17,"props":1565,"children":1566},{},[1567,1569,1574,1575,1580,1582,1587],{"type":21,"value":1568},"Yes, through ",{"type":16,"tag":24,"props":1570,"children":1572},{"className":1571},[],[1573],{"type":21,"value":1270},{"type":21,"value":47},{"type":16,"tag":24,"props":1576,"children":1578},{"className":1577},[],[1579],{"type":21,"value":1285},{"type":21,"value":1581}," variable. That token doesn't work with ",{"type":16,"tag":24,"props":1583,"children":1585},{"className":1584},[],[1586],{"type":21,"value":894},{"type":21,"value":1588},", which needs an API key.",{"type":16,"tag":1468,"props":1590,"children":1592},{"id":1591},"do-skills-and-claudemd-load-in-headless-mode",[1593],{"type":21,"value":1594},"Do skills and CLAUDE.md load in headless mode?",{"type":16,"tag":17,"props":1596,"children":1597},{},[1598,1600,1605,1607,1612],{"type":21,"value":1599},"Yes, unless you pass ",{"type":16,"tag":24,"props":1601,"children":1603},{"className":1602},[],[1604],{"type":21,"value":894},{"type":21,"value":1606},". User-invoked skills work when the prompt starts with ",{"type":16,"tag":24,"props":1608,"children":1610},{"className":1609},[],[1611],{"type":21,"value":1176},{"type":21,"value":386},{"title":7,"searchDepth":1614,"depth":1614,"links":1615},2,[1616,1617,1618,1619,1620,1621,1622,1623,1624],{"id":60,"depth":1614,"text":63},{"id":196,"depth":1614,"text":199},{"id":461,"depth":1614,"text":464},{"id":831,"depth":1614,"text":834},{"id":1068,"depth":1614,"text":1071},{"id":1204,"depth":1614,"text":1207},{"id":1295,"depth":1614,"text":1298},{"id":1417,"depth":1614,"text":1420},{"id":1463,"depth":1614,"text":1466,"children":1625},[1626,1628,1629,1630,1631],{"id":1470,"depth":1627,"text":1473},3,{"id":1488,"depth":1627,"text":1491},{"id":1519,"depth":1627,"text":1522},{"id":1560,"depth":1627,"text":1563},{"id":1591,"depth":1627,"text":1594},"markdown","content:posts:claude-code-headless.md","content","posts/claude-code-headless.md","md",{"loc":4},[1639,1643,1647,1651,1655,1658,1662,1666,1670,1674,1678,1682,1686,1690,1694,1698,1702,1706,1709,1713,1717,1721,1726,1730,1734],{"_path":1640,"title":1641,"description":1642,"lang":222},"/posts/api-nedir","API nedir? REST API, HTTP istekleri ve durum kodları","API nedir, ne işe yarar? curl ile gerçek bir API isteği, REST API nedir, HTTP metodları, durum kodları ve API'lerde kimlik doğrulama, sade bir dille.",{"_path":1644,"title":1645,"description":1646,"lang":222},"/posts/backend-nedir","Backend nedir, backend developer ne iş yapar? Frontend ile farkı","Backend nedir, backend developer ne iş yapar? Sunucu, API ve veritabanı tarafını, frontend ile farkını ve küçük bir NestJS örneğini anlatıyorum.",{"_path":1648,"title":1649,"description":1650,"lang":222},"/posts/bursa-gunubirlik-gezi-mudanya-koza-han-golyazi","Bursa günübirlik gezi: Mudanya, Koza Han, İskender ve Gölyazı","İstanbul'dan Bursa'ya günübirlik gezi: Mudanya'da kahvaltı, Koza Han, Mavi Dükkan'da Bursa iskender ve Gölyazı'daki 725 yıllık Ağlayan Çınar.",{"_path":1652,"title":1653,"description":1654,"lang":222},"/posts/claude-code-nedir","Claude Code nedir, nasıl kurulur ve kullanılır?","Claude Code nedir? Anthropic'in kodlama aracının kurulumu, giriş, temel komutlar, CLAUDE.md, izin modları ve sunucuda paralel kullanımı, örneklerle.",{"_path":1196,"title":1656,"description":1657},"Claude Code skills: what they are and how to write one","Claude Code skills explained: SKILL.md and its frontmatter, where skills live, how they trigger, and how to write one step by step, with a real example.",{"_path":1659,"title":1660,"description":1661,"lang":222},"/posts/docker-nedir","Docker nedir? Image, container, kurulum ve Docker Compose","Docker nedir, ne işe yarar? Image, container, volume ve network kavramları, Ubuntu'ya Docker kurulumu ve küçük bir docker-compose.yml örneği.",{"_path":1663,"title":1664,"description":1665,"lang":222},"/posts/full-stack-developer-nedir","Full stack developer nedir, ne iş yapar, nasıl olunur?","Full stack developer nedir, ne iş yapar? Frontend, backend, veritabanı ve deploy katmanları, gerçek bir stack, 2026 maaş verisi ve nasıl olunacağı.",{"_path":1667,"title":1668,"description":1669,"lang":222},"/posts/istanbul-yazilim-sirketleri","İstanbul yazılım şirketleri: nasıl seçilir ve kimler var? (2026)","İstanbul yazılım şirketleri arasında nasıl seçim yapılır? Sorulacak sorular, sözleşme ve kod sahipliği, uyarı işaretleri ve bilinen şirketlerden bir liste.",{"_path":1671,"title":1672,"description":1673},"/posts/nestjs-vs-express","NestJS vs Express: the same endpoint in both, and when to pick which","NestJS vs Express from a NestJS developer: the same API in both, testing, performance numbers from the official docs, and when to pick each.",{"_path":1675,"title":1676,"description":1677,"lang":222},"/posts/node-js-nedir","Node.js nedir? Event loop, npm ve ilk HTTP sunucusu","Node.js nedir, ne işe yarar? Event loop'un sade bir anlatımı, npm, sıfırdan bir HTTP sunucusu ve Node.js'in diğer backend seçenekleri arasındaki yeri.",{"_path":1679,"title":1680,"description":1681},"/posts/nuxt-seo","Nuxt SEO: the setup I use on this site, with the real config","A practical Nuxt SEO guide from halit.org: sitemap, robots, schema.org, useSeoMeta, real 404s, prerendering, hreflang and the nofollow default in Nuxt Content.",{"_path":1683,"title":1684,"description":1685},"/posts/nuxt-vs-next","Nuxt vs Next: an honest comparison from someone who ships Nuxt","Nuxt vs Next compared on Nuxt 4.6 and Next.js 16: rendering modes, routing, data fetching, ecosystem and deployment, with code for the same page in both.",{"_path":1687,"title":1688,"description":1689,"lang":222},"/posts/prometheus-grafana-nedir","Prometheus nedir, Grafana nedir? node_exporter ile sunucu izleme","Prometheus nedir, Grafana nedir ve birlikte nasıl çalışırlar? Metrikler, exporter'lar, node_exporter, PromQL örnekleri ve Docker Compose ile kurulum.",{"_path":1691,"title":1692,"description":1693},"/posts/teleparty-alternative","Teleparty alternative: free ways to watch together online","Looking for a Teleparty alternative? An honest look at Watchbear, Discord Watch Together and Watch2Gether for watching YouTube and Netflix with friends.",{"_path":1695,"title":1696,"description":1697,"lang":222},"/posts/typescript-nedir","TypeScript nedir? JavaScript'e ne ekler, tsconfig nasıl ayarlanır?","TypeScript nedir, JavaScript'ten farkı ne? Tipler, union, generic ve narrowing örnekleri, tsconfig.json'un temel ayarları ve Node.js ile kullanımı.",{"_path":1699,"title":1700,"description":1701,"lang":222},"/posts/yazilimci-maaslari-2026","Yazılımcı maaşları 2026: kıdem, pozisyon, şehir ve çalışma düzenine göre","Yazılımcı maaşları 2026: 5.002 kişilik anketin ham verisinden hesapladığım junior, middle, senior, pozisyon, şehir ve remote maaş ortancaları.",{"_path":1703,"title":1704,"description":1705,"lang":222},"/posts/yazilimci-nasil-olunur","Yazılımcı nasıl olunur? Yazılım öğrenmek için pratik bir yol","Yazılımcı nasıl olunur, yazılım öğrenmeye nereden başlanır? 2012'de forumlarla başlayıp 2020'den beri profesyonel çalışan bir yazılımcının pratik planı.",{"_path":180,"title":1707,"description":1708},"How I run multiple Claude Code agents in parallel on a server","My Claude Code setup: an Ubuntu server set up with claude-code-vps, parallel agents in a tmux grid with claudegrid, and the HaltOS desktop on top.",{"_path":1710,"title":1711,"description":1712},"/posts/small-web-tools-json-to-excel-steam-deck-watchbear","Small web tools I made: JSON to Excel, Steam Deck web apps, Watchbear","A JSON to Excel converter, a generator that puts any website in Steam Deck Game Mode, and Watchbear for watching videos in sync with friends.",{"_path":1714,"title":1715,"description":1716,"lang":222},"/posts/udf-dosyasi-nasil-acilir-pdf-word-cevirme","UDF dosyası nasıl açılır, PDF'ye ve Word'e nasıl çevrilir?","UYAP'ın UDF dosyasını program kurmadan tarayıcıda açmak, PDF'ye ve Word'e çevirmek, Word ve PDF'den UDF hazırlamak. Her aracın neyi koruyup neyi kaybettiğiyle birlikte.",{"_path":1718,"title":1719,"description":1720},"/posts/i-built-a-free-chrome-extension-to-watch-any-video-in-sync-with-friends","I built a free Chrome extension to watch any video in sync with friends","Watchbear is my free Chrome extension to watch YouTube together with friends: start a room, share the link, and play, pause and seek stay in sync.",{"_path":1722,"title":1723,"description":1724,"lang":1725},"/posts/a-short-trip-to-bursa","A day trip to Bursa: Mudanya, Koza Han, Iskender and Gölyazı","A day trip to Bursa via the Osmangazi Bridge, breakfast in Mudanya, Koza Han, Mavi Dükkan Iskender, and a peaceful visit to Gölyazı.","en",{"_path":1727,"title":1728,"description":1729},"/posts/embedding-metabase-dashboard-on-nuxt-app","Embedding a Metabase dashboard in a Nuxt 3 app","Metabase embedding in Nuxt 3: sign the embed JWT in a server route, lock parameters per user, set token expiry and theme the embedded dashboard.",{"_path":1731,"title":1732,"description":1733},"/posts/grafana-deep-dive","Grafana Docker Compose setup with Prometheus and node_exporter","Monitor a server with Grafana in Docker: a docker-compose.yml for Grafana, Prometheus and node_exporter, the scrape config, and a ready-made dashboard.",{"_path":1735,"title":1736,"description":1737},"/posts/deploy-express-app-on-vercel","Deploying an Express.js application on Vercel","How to deploy an Express app on Vercel: the zero-config way, the older vercel.json setup, environment variables, logs and serverless limits.",null,1795295806509]