[{"data":1,"prerenderedAt":1474},["ShallowReactive",2],{"blog-2026-02-05-claude-code-agent-teams-guide":3},{"id":4,"title":5,"author":6,"body":7,"date":1456,"description":1457,"extension":1458,"language":1459,"meta":1460,"navigation":1462,"path":1463,"seo":1464,"stem":1465,"tags":1466,"__hash__":1473},"blog/blog/2026-02-05-claude-code-agent-teams-guide.md","Claude Code Agent Teams: The Complete Guide to Multi-Agent Development","NeoAI",{"type":8,"value":9,"toc":1401},"minimark",[10,14,23,26,29,32,37,48,55,74,85,87,91,98,103,163,167,193,197,200,281,287,289,293,297,300,358,361,383,387,390,396,399,417,421,424,430,448,453,467,495,498,515,517,521,525,528,534,538,541,547,550,554,561,576,580,583,600,606,610,616,619,623,626,632,635,637,641,645,698,702,708,712,725,728,739,743,746,760,763,767,774,782,785,787,791,795,798,804,807,811,814,820,823,827,833,836,838,842,846,853,859,863,901,904,908,911,917,921,924,928,931,942,945,947,951,955,976,980,983,1027,1031,1037,1043,1047,1082,1084,1088,1091,1172,1174,1178,1185,1188,1191,1193,1197,1201,1228,1232,1238,1242,1311,1315,1338,1340,1344,1373,1381,1384,1386,1397],[11,12,5],"h1",{"id":13},"claude-code-agent-teams-the-complete-guide-to-multi-agent-development",[15,16,17,18,22],"p",{},"The era of single-agent AI assistance is ending. Anthropic just shipped one of the most requested features for Claude Code: ",[19,20,21],"strong",{},"Agent Teams"," — the ability to spawn, coordinate, and orchestrate multiple Claude instances working together on complex tasks.",[15,24,25],{},"This isn't just \"run multiple terminals.\" It's a fundamentally new paradigm where AI agents communicate with each other, claim tasks from shared queues, challenge each other's conclusions, and coordinate work without constant human supervision.",[15,27,28],{},"In this guide, we'll cover everything from basic setup to advanced patterns, with real examples you can use today.",[30,31],"hr",{},[33,34,36],"h2",{"id":35},"what-are-agent-teams","What Are Agent Teams?",[15,38,39,40,43,44,47],{},"Agent Teams let you coordinate multiple Claude Code instances working in parallel. One session acts as the ",[19,41,42],{},"team lead",", creating teammates, assigning work, and synthesizing results. ",[19,45,46],{},"Teammates"," are independent Claude instances, each with their own context window, working on their assigned tasks.",[15,49,50,51,54],{},"The key differentiator from Anthropic's existing subagents: ",[19,52,53],{},"teammates can talk to each other directly",". They're not just workers reporting back to a central coordinator — they're collaborators that can:",[56,57,58,62,65,68,71],"ul",{},[59,60,61],"li",{},"Message specific teammates or broadcast to the entire team",[59,63,64],{},"Claim and complete tasks from a shared task list",[59,66,67],{},"Challenge each other's findings (critical for debugging)",[59,69,70],{},"Request plan approval before major changes",[59,72,73],{},"Shut down gracefully when their work is done",[75,76,81],"pre",{"className":77,"code":79,"language":80},[78],"language-text","┌──────────────────────────────────────────────────────────────┐\n│                         YOU                                  │\n│                          │                                   │\n│                    ┌─────▼─────┐                             │\n│                    │   LEAD    │ ◄── Coordinates work        │\n│                    │  SESSION  │     Synthesizes results     │\n│                    └─────┬─────┘                             │\n│                          │                                   │\n│          ┌───────────────┼───────────────┐                   │\n│          │               │               │                   │\n│    ┌─────▼─────┐   ┌─────▼─────┐   ┌─────▼─────┐             │\n│    │ TEAMMATE  │◄─►│ TEAMMATE  │◄─►│ TEAMMATE  │             │\n│    │  (coder)  │   │(reviewer) │   │ (tester)  │             │\n│    └───────────┘   └───────────┘   └───────────┘             │\n│          │               │               │                   │\n│          └───────────────┼───────────────┘                   │\n│                          │                                   │\n│                    ┌─────▼─────┐                             │\n│                    │   TASK    │ ◄── Shared work queue       │\n│                    │   LIST    │     with dependencies       │\n│                    └───────────┘                             │\n└──────────────────────────────────────────────────────────────┘\n","text",[82,83,79],"code",{"__ignoreMap":84},"",[30,86],{},[33,88,90],{"id":89},"when-should-you-use-agent-teams","When Should You Use Agent Teams?",[15,92,93,94,97],{},"Agent Teams shine when ",[19,95,96],{},"parallel exploration adds value",". They're not a silver bullet — coordination overhead and token costs are significant. Here's when they make sense:",[99,100,102],"h3",{"id":101},"perfect-use-cases","✅ Perfect Use Cases",[104,105,106,119],"table",{},[107,108,109],"thead",{},[110,111,112,116],"tr",{},[113,114,115],"th",{},"Scenario",[113,117,118],{},"Why Teams Excel",[120,121,122,133,143,153],"tbody",{},[110,123,124,130],{},[125,126,127],"td",{},[19,128,129],{},"Research & Review",[125,131,132],{},"Multiple perspectives simultaneously — security, performance, correctness",[110,134,135,140],{},[125,136,137],{},[19,138,139],{},"New Modules/Features",[125,141,142],{},"Each teammate owns a separate piece, no stepping on toes",[110,144,145,150],{},[125,146,147],{},[19,148,149],{},"Debugging with Hypotheses",[125,151,152],{},"Competing theories tested in parallel, faster convergence",[110,154,155,160],{},[125,156,157],{},[19,158,159],{},"Cross-Layer Changes",[125,161,162],{},"Frontend, backend, tests each handled by specialists",[99,164,166],{"id":165},"when-to-avoid","❌ When to Avoid",[56,168,169,175,181,187],{},[59,170,171,174],{},[19,172,173],{},"Sequential tasks"," — One thing must finish before the next",[59,176,177,180],{},[19,178,179],{},"Same-file edits"," — Two teammates editing one file = merge hell",[59,182,183,186],{},[19,184,185],{},"Heavy dependencies"," — Tasks that constantly block each other",[59,188,189,192],{},[19,190,191],{},"Simple tasks"," — Overhead exceeds benefit",[99,194,196],{"id":195},"agent-teams-vs-subagents","Agent Teams vs. Subagents",[15,198,199],{},"Both parallelize work, but they're fundamentally different:",[104,201,202,214],{},[107,203,204],{},[110,205,206,209,212],{},[113,207,208],{},"Aspect",[113,210,211],{},"Subagents",[113,213,21],{},[120,215,216,229,242,255,268],{},[110,217,218,223,226],{},[125,219,220],{},[19,221,222],{},"Context",[125,224,225],{},"Own context, results return to caller",[125,227,228],{},"Own context, fully independent",[110,230,231,236,239],{},[125,232,233],{},[19,234,235],{},"Communication",[125,237,238],{},"Report to main agent only",[125,240,241],{},"Message each other directly",[110,243,244,249,252],{},[125,245,246],{},[19,247,248],{},"Coordination",[125,250,251],{},"Main agent manages all work",[125,253,254],{},"Shared task list, self-coordination",[110,256,257,262,265],{},[125,258,259],{},[19,260,261],{},"Best For",[125,263,264],{},"Focused tasks, only result matters",[125,266,267],{},"Complex work requiring collaboration",[110,269,270,275,278],{},[125,271,272],{},[19,273,274],{},"Token Cost",[125,276,277],{},"Lower (summarized results)",[125,279,280],{},"Higher (each is a full Claude instance)",[15,282,283,286],{},[19,284,285],{},"Rule of thumb:"," Use subagents for quick, focused helpers. Use agent teams when workers need to discuss, debate, and coordinate.",[30,288],{},[33,290,292],{"id":291},"getting-started","Getting Started",[99,294,296],{"id":295},"_1-enable-agent-teams","1. Enable Agent Teams",[15,298,299],{},"Agent teams are experimental and disabled by default. Enable them via environment variable or settings:",[75,301,305],{"className":302,"code":303,"language":304,"meta":84,"style":84},"language-json shiki shiki-themes github-light github-dark","// ~/.claude/settings.json\n{\n  \"env\": {\n    \"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS\": \"1\"\n  }\n}\n","json",[82,306,307,316,323,333,346,352],{"__ignoreMap":84},[308,309,312],"span",{"class":310,"line":311},"line",1,[308,313,315],{"class":314},"sJ8bj","// ~/.claude/settings.json\n",[308,317,319],{"class":310,"line":318},2,[308,320,322],{"class":321},"sVt8B","{\n",[308,324,326,330],{"class":310,"line":325},3,[308,327,329],{"class":328},"sj4cs","  \"env\"",[308,331,332],{"class":321},": {\n",[308,334,336,339,342],{"class":310,"line":335},4,[308,337,338],{"class":328},"    \"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS\"",[308,340,341],{"class":321},": ",[308,343,345],{"class":344},"sZZnC","\"1\"\n",[308,347,349],{"class":310,"line":348},5,[308,350,351],{"class":321},"  }\n",[308,353,355],{"class":310,"line":354},6,[308,356,357],{"class":321},"}\n",[15,359,360],{},"Or export in your shell:",[75,362,366],{"className":363,"code":364,"language":365,"meta":84,"style":84},"language-bash shiki shiki-themes github-light github-dark","export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1\n","bash",[82,367,368],{"__ignoreMap":84},[308,369,370,374,377,380],{"class":310,"line":311},[308,371,373],{"class":372},"szBVR","export",[308,375,376],{"class":321}," CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS",[308,378,379],{"class":372},"=",[308,381,382],{"class":328},"1\n",[99,384,386],{"id":385},"_2-create-your-first-team","2. Create Your First Team",[15,388,389],{},"Just describe what you want in natural language. Claude handles the rest:",[75,391,394],{"className":392,"code":393,"language":80},[78],"I'm building a CLI tool for tracking TODO comments in codebases. \nCreate an agent team to explore this from different angles:\n- One teammate on UX design\n- One on technical architecture  \n- One playing devil's advocate, finding holes in our approach\n",[82,395,393],{"__ignoreMap":84},[15,397,398],{},"Claude will:",[400,401,402,405,408,411,414],"ol",{},[59,403,404],{},"Create a team with a shared task list",[59,406,407],{},"Spawn three teammates with those roles",[59,409,410],{},"Have them explore the problem independently",[59,412,413],{},"Synthesize findings into recommendations",[59,415,416],{},"Clean up the team when finished",[99,418,420],{"id":419},"_3-choose-a-display-mode","3. Choose a Display Mode",[15,422,423],{},"Two options for seeing your team work:",[15,425,426,429],{},[19,427,428],{},"In-Process Mode"," (default)",[56,431,432,435,442,445],{},[59,433,434],{},"All teammates run in your main terminal",[59,436,437,438,441],{},"Use ",[82,439,440],{},"Shift+Up/Down"," to select teammates",[59,443,444],{},"Type to message them directly",[59,446,447],{},"Works in any terminal",[15,449,450],{},[19,451,452],{},"Split-Pane Mode",[56,454,455,458,461,464],{},[59,456,457],{},"Each teammate gets its own pane",[59,459,460],{},"See everyone's output simultaneously",[59,462,463],{},"Click into any pane to interact",[59,465,466],{},"Requires tmux or iTerm2",[75,468,470],{"className":302,"code":469,"language":304,"meta":84,"style":84},"// Force split-pane mode\n{\n  \"teammateMode\": \"tmux\"\n}\n",[82,471,472,477,481,491],{"__ignoreMap":84},[308,473,474],{"class":310,"line":311},[308,475,476],{"class":314},"// Force split-pane mode\n",[308,478,479],{"class":310,"line":318},[308,480,322],{"class":321},[308,482,483,486,488],{"class":310,"line":325},[308,484,485],{"class":328},"  \"teammateMode\"",[308,487,341],{"class":321},[308,489,490],{"class":344},"\"tmux\"\n",[308,492,493],{"class":310,"line":335},[308,494,357],{"class":321},[15,496,497],{},"Or per-session:",[75,499,501],{"className":363,"code":500,"language":365,"meta":84,"style":84},"claude --teammate-mode in-process\n",[82,502,503],{"__ignoreMap":84},[308,504,505,509,512],{"class":310,"line":311},[308,506,508],{"class":507},"sScJk","claude",[308,510,511],{"class":328}," --teammate-mode",[308,513,514],{"class":344}," in-process\n",[30,516],{},[33,518,520],{"id":519},"controlling-your-team","Controlling Your Team",[99,522,524],{"id":523},"assign-tasks","Assign Tasks",[15,526,527],{},"The lead coordinates based on your natural language instructions:",[75,529,532],{"className":530,"code":531,"language":80},[78],"Create 4 teammates to refactor these modules in parallel.\nUse Sonnet for each one.\n",[82,533,531],{"__ignoreMap":84},[99,535,537],{"id":536},"require-plan-approval","Require Plan Approval",[15,539,540],{},"For risky changes, make teammates plan before implementing:",[75,542,545],{"className":543,"code":544,"language":80},[78],"Spawn an architect teammate to refactor authentication.\nRequire plan approval before they make any changes.\n",[82,546,544],{"__ignoreMap":84},[15,548,549],{},"The teammate enters read-only \"plan mode,\" proposes an approach, and waits for lead approval before proceeding. If rejected, they revise and resubmit.",[99,551,553],{"id":552},"use-delegate-mode","Use Delegate Mode",[15,555,556,557,560],{},"Sometimes the lead starts coding instead of delegating. ",[19,558,559],{},"Delegate mode"," restricts the lead to coordination-only tools:",[400,562,563,566,573],{},[59,564,565],{},"Start a team first",[59,567,568,569,572],{},"Press ",[82,570,571],{},"Shift+Tab"," to enter delegate mode",[59,574,575],{},"Lead can only: spawn, message, shut down, manage tasks",[99,577,579],{"id":578},"talk-directly-to-teammates","Talk Directly to Teammates",[15,581,582],{},"Each teammate is a full Claude Code session. Message them anytime:",[56,584,585,594],{},[59,586,587,590,591,593],{},[19,588,589],{},"In-process:"," ",[82,592,440],{}," to select, then type",[59,595,596,599],{},[19,597,598],{},"Split-pane:"," Click into their pane",[75,601,604],{"className":602,"code":603,"language":80},[78],"Hey security reviewer, also check for SQL injection in the user input handlers\n",[82,605,603],{"__ignoreMap":84},[99,607,609],{"id":608},"shut-down-teammates","Shut Down Teammates",[75,611,614],{"className":612,"code":613,"language":80},[78],"Ask the researcher teammate to shut down\n",[82,615,613],{"__ignoreMap":84},[15,617,618],{},"The lead sends a shutdown request. The teammate can approve (exit gracefully) or reject with an explanation if they're mid-task.",[99,620,622],{"id":621},"clean-up","Clean Up",[15,624,625],{},"When done:",[75,627,630],{"className":628,"code":629,"language":80},[78],"Clean up the team\n",[82,631,629],{"__ignoreMap":84},[15,633,634],{},"This removes shared resources. Always shut down teammates first — cleanup fails if any are still running.",[30,636],{},[33,638,640],{"id":639},"how-it-works-under-the-hood","How It Works Under the Hood",[99,642,644],{"id":643},"architecture","Architecture",[104,646,647,657],{},[107,648,649],{},[110,650,651,654],{},[113,652,653],{},"Component",[113,655,656],{},"Role",[120,658,659,669,678,688],{},[110,660,661,666],{},[125,662,663],{},[19,664,665],{},"Team Lead",[125,667,668],{},"Main session that creates the team and coordinates",[110,670,671,675],{},[125,672,673],{},[19,674,46],{},[125,676,677],{},"Independent Claude instances with assigned tasks",[110,679,680,685],{},[125,681,682],{},[19,683,684],{},"Task List",[125,686,687],{},"Shared work queue with status tracking",[110,689,690,695],{},[125,691,692],{},[19,693,694],{},"Mailbox",[125,696,697],{},"Inter-agent messaging system",[99,699,701],{"id":700},"storage-locations","Storage Locations",[75,703,706],{"className":704,"code":705,"language":80},[78],"~/.claude/\n├── teams/{team-name}/\n│   └── config.json          # Team metadata, member list\n└── tasks/{team-name}/       # Shared task queue\n",[82,707,705],{"__ignoreMap":84},[99,709,711],{"id":710},"task-states-dependencies","Task States & Dependencies",[15,713,714,715,718,719,718,722],{},"Tasks flow through: ",[82,716,717],{},"pending"," → ",[82,720,721],{},"in progress",[82,723,724],{},"completed",[15,726,727],{},"Dependencies work automatically. If Task B depends on Task A:",[56,729,730,736],{},[59,731,732,733,735],{},"Task B stays ",[82,734,717],{}," until Task A completes",[59,737,738],{},"When A finishes, B unblocks without manual intervention",[99,740,742],{"id":741},"claiming-tasks","Claiming Tasks",[15,744,745],{},"Two modes:",[400,747,748,754],{},[59,749,750,753],{},[19,751,752],{},"Lead assigns:"," Explicitly tell lead who does what",[59,755,756,759],{},[19,757,758],{},"Self-claim:"," Teammates grab the next unassigned, unblocked task",[15,761,762],{},"File locking prevents race conditions when multiple teammates reach for the same task.",[99,764,766],{"id":765},"token-reality-check","Token Reality Check",[15,768,769,770,773],{},"Agent teams are ",[19,771,772],{},"expensive",". Each teammate has their own context window. Anthropic's data shows:",[56,775,776,779],{},[59,777,778],{},"Single agent: baseline",[59,780,781],{},"Agent teams: ~15× more tokens than chat",[15,783,784],{},"This is worth it for complex research, parallel review, and feature development. For routine tasks, stick to single sessions or subagents.",[30,786],{},[33,788,790],{"id":789},"real-world-examples","Real-World Examples",[99,792,794],{"id":793},"_1-parallel-code-review","1. Parallel Code Review",[15,796,797],{},"A single reviewer tends to focus on one issue type at a time. Split the concerns:",[75,799,802],{"className":800,"code":801,"language":80},[78],"Create an agent team to review PR #142. Spawn three reviewers:\n- One focused on security implications\n- One checking performance impact  \n- One validating test coverage\nHave them each review and report findings.\n",[82,803,801],{"__ignoreMap":84},[15,805,806],{},"Each reviewer applies a different lens to the same PR. The lead synthesizes across all three.",[99,808,810],{"id":809},"_2-debugging-with-competing-hypotheses","2. Debugging with Competing Hypotheses",[15,812,813],{},"When root cause is unclear, single agents anchor on their first theory. Fight this with adversarial investigation:",[75,815,818],{"className":816,"code":817,"language":80},[78],"Users report the app exits after one message instead of staying connected.\nSpawn 5 teammates to investigate different hypotheses. Have them talk to\neach other to try to disprove each other's theories, like a scientific\ndebate. Update the findings doc with whatever consensus emerges.\n",[82,819,817],{"__ignoreMap":84},[15,821,822],{},"The \"scientific debate\" structure forces teammates to actively challenge each other. The surviving theory is far more likely to be the actual root cause.",[99,824,826],{"id":825},"_3-cross-layer-feature-implementation","3. Cross-Layer Feature Implementation",[75,828,831],{"className":829,"code":830,"language":80},[78],"Build OAuth login. Create a team with:\n- Backend developer: implement OAuth controller and models\n- Frontend developer: build login UI components\n- Test writer: create integration tests (depends on backend)\nHave them coordinate via the task list.\n",[82,832,830],{"__ignoreMap":84},[15,834,835],{},"Each teammate owns their layer. The test writer's tasks block until backend is ready.",[30,837],{},[33,839,841],{"id":840},"best-practices","Best Practices",[99,843,845],{"id":844},"give-teammates-rich-context","Give Teammates Rich Context",[15,847,848,849,852],{},"Teammates load project context (CLAUDE.md, MCP servers, skills) but ",[19,850,851],{},"not the lead's conversation history",". Include everything they need in the spawn prompt:",[75,854,857],{"className":855,"code":856,"language":80},[78],"Spawn a security reviewer with the prompt: \"Review the auth module at \nsrc/auth/ for vulnerabilities. Focus on token handling, session management,\nand input validation. The app uses JWT in httpOnly cookies. Report issues\nwith severity ratings.\"\n",[82,858,856],{"__ignoreMap":84},[99,860,862],{"id":861},"size-tasks-appropriately","Size Tasks Appropriately",[104,864,865,875],{},[107,866,867],{},[110,868,869,872],{},[113,870,871],{},"Size",[113,873,874],{},"Problem",[120,876,877,885,893],{},[110,878,879,882],{},[125,880,881],{},"Too small",[125,883,884],{},"Coordination overhead exceeds benefit",[110,886,887,890],{},[125,888,889],{},"Too large",[125,891,892],{},"Teammates work too long without check-ins",[110,894,895,898],{},[125,896,897],{},"Just right",[125,899,900],{},"Self-contained units with clear deliverables",[15,902,903],{},"Good task examples: a function, a test file, a review of one module.",[99,905,907],{"id":906},"prevent-file-conflicts","Prevent File Conflicts",[15,909,910],{},"Two teammates editing the same file = disaster. Break work so each owns different files:",[75,912,915],{"className":913,"code":914,"language":80},[78],"Teammate A owns: src/auth/*\nTeammate B owns: src/api/*\nTeammate C owns: tests/*\n",[82,916,914],{"__ignoreMap":84},[99,918,920],{"id":919},"monitor-and-steer","Monitor and Steer",[15,922,923],{},"Don't let the team run unattended for too long. Check progress, redirect bad approaches, synthesize findings as they arrive.",[99,925,927],{"id":926},"start-with-research","Start with Research",[15,929,930],{},"If you're new to agent teams, begin with non-coding tasks:",[56,932,933,936,939],{},[59,934,935],{},"Reviewing a PR",[59,937,938],{},"Researching a library",[59,940,941],{},"Investigating a bug",[15,943,944],{},"These show the value of parallel exploration without the coordination complexity of parallel implementation.",[30,946],{},[33,948,950],{"id":949},"troubleshooting","Troubleshooting",[99,952,954],{"id":953},"teammates-not-appearing","Teammates Not Appearing",[400,956,957,964,967,973],{},[59,958,959,960,963],{},"In in-process mode, press ",[82,961,962],{},"Shift+Down"," — they might be running invisibly",[59,965,966],{},"Check if Claude thought the task was too simple for a team",[59,968,969,970],{},"For split panes, verify tmux is installed: ",[82,971,972],{},"which tmux",[59,974,975],{},"For iTerm2, ensure Python API is enabled in preferences",[99,977,979],{"id":978},"too-many-permission-prompts","Too Many Permission Prompts",[15,981,982],{},"Teammate permissions bubble up to the lead, causing friction. Pre-approve common operations:",[75,984,986],{"className":302,"code":985,"language":304,"meta":84,"style":84},"{\n  \"permissions\": {\n    \"allow\": [\"Write(*)\", \"Bash(npm test)\"]\n  }\n}\n",[82,987,988,992,999,1019,1023],{"__ignoreMap":84},[308,989,990],{"class":310,"line":311},[308,991,322],{"class":321},[308,993,994,997],{"class":310,"line":318},[308,995,996],{"class":328},"  \"permissions\"",[308,998,332],{"class":321},[308,1000,1001,1004,1007,1010,1013,1016],{"class":310,"line":325},[308,1002,1003],{"class":328},"    \"allow\"",[308,1005,1006],{"class":321},": [",[308,1008,1009],{"class":344},"\"Write(*)\"",[308,1011,1012],{"class":321},", ",[308,1014,1015],{"class":344},"\"Bash(npm test)\"",[308,1017,1018],{"class":321},"]\n",[308,1020,1021],{"class":310,"line":335},[308,1022,351],{"class":321},[308,1024,1025],{"class":310,"line":348},[308,1026,357],{"class":321},[99,1028,1030],{"id":1029},"lead-doing-work-instead-of-delegating","Lead Doing Work Instead of Delegating",[75,1032,1035],{"className":1033,"code":1034,"language":80},[78],"Wait for your teammates to complete their tasks before proceeding\n",[82,1036,1034],{"__ignoreMap":84},[15,1038,1039,1040,1042],{},"Or enable delegate mode (",[82,1041,571],{},").",[99,1044,1046],{"id":1045},"orphaned-tmux-sessions","Orphaned tmux Sessions",[75,1048,1050],{"className":363,"code":1049,"language":365,"meta":84,"style":84},"tmux ls\ntmux kill-session -t \u003Csession-name>\n",[82,1051,1052,1060],{"__ignoreMap":84},[308,1053,1054,1057],{"class":310,"line":311},[308,1055,1056],{"class":507},"tmux",[308,1058,1059],{"class":344}," ls\n",[308,1061,1062,1064,1067,1070,1073,1076,1079],{"class":310,"line":318},[308,1063,1056],{"class":507},[308,1065,1066],{"class":344}," kill-session",[308,1068,1069],{"class":328}," -t",[308,1071,1072],{"class":372}," \u003C",[308,1074,1075],{"class":344},"session-nam",[308,1077,1078],{"class":321},"e",[308,1080,1081],{"class":372},">\n",[30,1083],{},[33,1085,1087],{"id":1086},"current-limitations","Current Limitations",[15,1089,1090],{},"Agent teams are still experimental. Key constraints:",[104,1092,1093,1103],{},[107,1094,1095],{},[110,1096,1097,1100],{},[113,1098,1099],{},"Limitation",[113,1101,1102],{},"Impact",[120,1104,1105,1116,1124,1132,1140,1148,1156,1164],{},[110,1106,1107,1110],{},[125,1108,1109],{},"No session resumption",[125,1111,1112,1115],{},[82,1113,1114],{},"/resume"," doesn't restore in-process teammates",[110,1117,1118,1121],{},[125,1119,1120],{},"Task status lag",[125,1122,1123],{},"Teammates sometimes forget to mark tasks complete",[110,1125,1126,1129],{},[125,1127,1128],{},"Slow shutdown",[125,1130,1131],{},"Teammates finish current tool call before exiting",[110,1133,1134,1137],{},[125,1135,1136],{},"One team per session",[125,1138,1139],{},"Clean up before starting a new team",[110,1141,1142,1145],{},[125,1143,1144],{},"No nested teams",[125,1146,1147],{},"Teammates can't spawn their own teams",[110,1149,1150,1153],{},[125,1151,1152],{},"Fixed lead",[125,1154,1155],{},"Can't promote a teammate to lead",[110,1157,1158,1161],{},[125,1159,1160],{},"Permissions at spawn",[125,1162,1163],{},"Can't set per-teammate modes upfront",[110,1165,1166,1169],{},[125,1167,1168],{},"Split panes limited",[125,1170,1171],{},"Not supported in VS Code terminal, Windows Terminal, Ghostty",[30,1173],{},[33,1175,1177],{"id":1176},"the-bigger-picture","The Bigger Picture",[15,1179,1180,1181,1184],{},"Gartner predicts that by end of 2026, ",[19,1182,1183],{},"40% of enterprise applications will include task-specific AI agents",". Multi-agent inquiry has surged 1,445% in just one year.",[15,1186,1187],{},"Agent Teams in Claude Code isn't just a feature — it's a preview of how software development is evolving. The pattern of specialized agents collaborating on complex tasks, each with their own expertise and context, mirrors how high-performing human teams operate.",[15,1189,1190],{},"Today it's experimental, behind a feature flag. Tomorrow it'll be how we build software.",[30,1192],{},[33,1194,1196],{"id":1195},"quick-reference","Quick Reference",[99,1198,1200],{"id":1199},"enable","Enable",[75,1202,1204],{"className":302,"code":1203,"language":304,"meta":84,"style":84},"{ \"env\": { \"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS\": \"1\" } }\n",[82,1205,1206],{"__ignoreMap":84},[308,1207,1208,1211,1214,1217,1220,1222,1225],{"class":310,"line":311},[308,1209,1210],{"class":321},"{ ",[308,1212,1213],{"class":328},"\"env\"",[308,1215,1216],{"class":321},": { ",[308,1218,1219],{"class":328},"\"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS\"",[308,1221,341],{"class":321},[308,1223,1224],{"class":344},"\"1\"",[308,1226,1227],{"class":321}," } }\n",[99,1229,1231],{"id":1230},"create-team","Create Team",[75,1233,1236],{"className":1234,"code":1235,"language":80},[78],"Create an agent team with [roles] to [task]\n",[82,1237,1235],{"__ignoreMap":84},[99,1239,1241],{"id":1240},"control","Control",[104,1243,1244,1254],{},[107,1245,1246],{},[110,1247,1248,1251],{},[113,1249,1250],{},"Action",[113,1252,1253],{},"Method",[120,1255,1256,1265,1275,1283,1291,1303],{},[110,1257,1258,1261],{},[125,1259,1260],{},"Select teammate",[125,1262,1263],{},[82,1264,440],{},[110,1266,1267,1270],{},[125,1268,1269],{},"View task list",[125,1271,1272],{},[82,1273,1274],{},"Ctrl+T",[110,1276,1277,1279],{},[125,1278,559],{},[125,1280,1281],{},[82,1282,571],{},[110,1284,1285,1288],{},[125,1286,1287],{},"Message teammate",[125,1289,1290],{},"Select + type",[110,1292,1293,1296],{},[125,1294,1295],{},"Shut down",[125,1297,1298,1299,1302],{},"\"Ask ",[308,1300,1301],{},"name"," to shut down\"",[110,1304,1305,1308],{},[125,1306,1307],{},"Clean up",[125,1309,1310],{},"\"Clean up the team\"",[99,1312,1314],{"id":1313},"display-modes","Display Modes",[75,1316,1318],{"className":302,"code":1317,"language":304,"meta":84,"style":84},"{ \"teammateMode\": \"in-process\" }  // or \"tmux\" / \"auto\"\n",[82,1319,1320],{"__ignoreMap":84},[308,1321,1322,1324,1327,1329,1332,1335],{"class":310,"line":311},[308,1323,1210],{"class":321},[308,1325,1326],{"class":328},"\"teammateMode\"",[308,1328,341],{"class":321},[308,1330,1331],{"class":344},"\"in-process\"",[308,1333,1334],{"class":321}," }  ",[308,1336,1337],{"class":314},"// or \"tmux\" / \"auto\"\n",[30,1339],{},[33,1341,1343],{"id":1342},"whats-next","What's Next",[56,1345,1346,1355,1364],{},[59,1347,1348,1354],{},[19,1349,1350],{},[1351,1352,211],"a",{"href":1353},"/docs/en/sub-agents"," — Lightweight delegation for focused tasks",[59,1356,1357,1363],{},[19,1358,1359],{},[1351,1360,1362],{"href":1361},"/docs/en/common-workflows","Git Worktrees"," — Manual parallel sessions",[59,1365,1366,1372],{},[19,1367,1368],{},[1351,1369,1371],{"href":1370},"/docs/en/mcp","MCP Servers"," — Extend agents with custom tools",[15,1374,1375,1376,1380],{},"The future of coding isn't you ",[1377,1378,1379],"em",{},"or"," AI. It's you orchestrating a team of AI specialists, each doing what they do best, while you focus on the big picture.",[15,1382,1383],{},"Welcome to multi-agent development.",[30,1385],{},[15,1387,1388],{},[1377,1389,1390,1391],{},"For the official documentation, visit ",[1351,1392,1396],{"href":1393,"rel":1394},"https://code.claude.com/docs/en/agent-teams",[1395],"nofollow","code.claude.com/docs/en/agent-teams",[1398,1399,1400],"style",{},"html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}",{"title":84,"searchDepth":318,"depth":318,"links":1402},[1403,1404,1409,1414,1422,1429,1434,1441,1447,1448,1449,1455],{"id":35,"depth":318,"text":36},{"id":89,"depth":318,"text":90,"children":1405},[1406,1407,1408],{"id":101,"depth":325,"text":102},{"id":165,"depth":325,"text":166},{"id":195,"depth":325,"text":196},{"id":291,"depth":318,"text":292,"children":1410},[1411,1412,1413],{"id":295,"depth":325,"text":296},{"id":385,"depth":325,"text":386},{"id":419,"depth":325,"text":420},{"id":519,"depth":318,"text":520,"children":1415},[1416,1417,1418,1419,1420,1421],{"id":523,"depth":325,"text":524},{"id":536,"depth":325,"text":537},{"id":552,"depth":325,"text":553},{"id":578,"depth":325,"text":579},{"id":608,"depth":325,"text":609},{"id":621,"depth":325,"text":622},{"id":639,"depth":318,"text":640,"children":1423},[1424,1425,1426,1427,1428],{"id":643,"depth":325,"text":644},{"id":700,"depth":325,"text":701},{"id":710,"depth":325,"text":711},{"id":741,"depth":325,"text":742},{"id":765,"depth":325,"text":766},{"id":789,"depth":318,"text":790,"children":1430},[1431,1432,1433],{"id":793,"depth":325,"text":794},{"id":809,"depth":325,"text":810},{"id":825,"depth":325,"text":826},{"id":840,"depth":318,"text":841,"children":1435},[1436,1437,1438,1439,1440],{"id":844,"depth":325,"text":845},{"id":861,"depth":325,"text":862},{"id":906,"depth":325,"text":907},{"id":919,"depth":325,"text":920},{"id":926,"depth":325,"text":927},{"id":949,"depth":318,"text":950,"children":1442},[1443,1444,1445,1446],{"id":953,"depth":325,"text":954},{"id":978,"depth":325,"text":979},{"id":1029,"depth":325,"text":1030},{"id":1045,"depth":325,"text":1046},{"id":1086,"depth":318,"text":1087},{"id":1176,"depth":318,"text":1177},{"id":1195,"depth":318,"text":1196,"children":1450},[1451,1452,1453,1454],{"id":1199,"depth":325,"text":1200},{"id":1230,"depth":325,"text":1231},{"id":1240,"depth":325,"text":1241},{"id":1313,"depth":325,"text":1314},{"id":1342,"depth":318,"text":1343},"2026-02-05","Everything you need to know about Claude Code's new Agent Teams feature — coordinating multiple AI instances that communicate, collaborate, and conquer complex development tasks together.","md",null,{"image":1461},"/blog/claude-code-agent-teams.png",true,"/blog/2026-02-05-claude-code-agent-teams-guide",{"title":5,"description":1457},"blog/2026-02-05-claude-code-agent-teams-guide",[1467,508,1468,1469,1470,1471,1472],"ai","agent-teams","multi-agent","development","orchestration","tutorial","KGIRwKVz3InZ4eJCnMDoVmGgXZ8oguDXy4K1JAi9BBQ",1784088102146]