Skip to content
Back to search

claude-council

SkillWorks with CodexGitHub

Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, Kimi, any model OpenRouter routes to, plus codex, antigravity, grok, kimi and cursor CLIs when installed, and a local ollama model) to get diverse perspectives on coding problems, with a streaming pane in tmux or inside Claude Code, and, once switched on, a heads-up side-check where one council seat, codex by default, rereads a long turn and shows one line above the prompt about something you may have missed.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/hex-claude-council/badge.svg)](/listings/hex-claude-council)

HTML

<a href="/listings/hex-claude-council"><img src="/api/listings/hex-claude-council/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

You need `curl` and `jq` on macOS, Linux, or Windows through Git Bash.

/plugin marketplace add hex/claude-marketplace
/plugin install claude-council

Give it at least one provider in the shell you start Claude Code from:

export OPENAI_API_KEY="..."
# Or GEMINI_API_KEY, XAI_API_KEY, PERPLEXITY_API_KEY,
# KIMI_API_KEY, OPENROUTER_API_KEY

Or store the keys in the plugin's config. Run `/plugin configure claude-council`, fill in the API key rows, and choose Save configuration. Claude Code keeps them in its secure storage, not in a settings file. The council's mod sets each stored key as its variable for the whole session, so every command Claude Code starts, in every project, gets the stored key in place of one your shell exports. The OpenRouter models row takes the model ids `OPENROUTER_MODELS` takes, so the router's seats need no export either. Stored keys need Claude Code 2.1.287 or later. The desktop app's plugin Settings tab takes them too, for sessions on that computer. They never reach claude.ai chat or Cowork. [API keys](docs/REFERENCE.md#api-keys) covers where they are kept, when the mod does not load, and how a cleared key behaves.

The `codex`, `agy` (Antigravity), `grok`, `kimi`, and `cursor-agent` CLIs use your existing subscriptions without an API key. Claude Code itself, the Copilot CLI and OpenCode can sit too, when named in `--providers`. A local `ollama` model needs no key or external network. With none of these, `--local` uses Claude subagents. See [provider selection](docs/REFERENCE.md#choosing-providers) and [local councils](docs/USAGE.md#local-council---local).

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/plugin.json · Manifest · 109 lines
  1. {
  2. "name": "claude-council",
  3. "version": "2026.11.2",
  4. "description": "Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, Kimi, any model OpenRouter routes to, plus codex, antigravity, grok, kimi and cursor CLIs when installed, and a local ollama model) to get diverse perspectives on coding problems, with a streaming pane in tmux or inside Claude Code, and, once switched on, a heads-up side-check where one council seat, codex by default, rereads a long turn and shows one line above the prompt about something you may have missed.",
  5. "author": {
  6. "name": "hex",
  7. "url": "https://github.com/hex"
  8. },
  9. "license": "MIT",
  10. "homepage": "https://github.com/hex/claude-council",
  11. "repository": "https://github.com/hex/claude-council",
  12. "icon": "./assets/icon.png",
  13. "types": "./mods/council-pane/types/index.d.ts",
  14. "keywords": [
  15. "multi-agent",
  16. "council",
  17. "second-opinion",
  18. "code-review",
  19. "consensus",
  20. "perspectives",
  21. "gemini",
  22. "openai",
  23. "grok",
  24. "perplexity",
  25. "kimi",
  26. "openrouter",
  27. "codex",
  28. "antigravity",
  29. "ollama",
  30. "cli",
  31. "pane",
  32. "tmux",
  33. "mod",
  34. "heads-up"
  35. ],
  36. "userConfig": {
  37. "pane_host": {
  38. "type": "string",
  39. "title": "Pane opens in",
  40. "description": "Where the council pane opens. claude-code draws it inside Claude Code; tmux leaves it to a tmux pane; ask asks once, the first time a run starts inside tmux, and remembers the answer (/council-pane ask forgets it). Outside tmux the pane opens here either way.",
  41. "default": "ask"
  42. },
  43. "collapse_when_done": {
  44. "type": "boolean",
  45. "title": "Collapse status when done",
  46. "description": "Open a finished run on its summary and the synthesis, with the seats folded below. Off keeps the live order: the seats, then the synthesis.",
  47. "default": true
  48. },
  49. "wake_on_async_done": {
  50. "type": "boolean",
  51. "title": "Wake on async finish",
  52. "description": "When a background council run (--async) finishes, submit a prompt so Claude fetches the result by itself. That starts a model turn and costs tokens; a failed run wakes nobody.",
  53. "default": false
  54. },
  55. "council_tool": {
  56. "type": "boolean",
  57. "title": "Council as a tool",
  58. "description": "Let Claude call the council as a tool, not only through /claude-council:ask. Every call shows you the question and the providers and asks before anything is sent.",
  59. "default": true
  60. },
  61. "heads_up": {
  62. "type": "boolean",
  63. "title": "Heads-up side-check",
  64. "description": "Turn the heads-up side-check on in every project: while Claude works a long turn, one council seat, codex by default, rereads the last five turns, the last eight tool results and the text Claude wrote with Write or Edit, and may show one line above the prompt. Nothing asks before a send; that material leaves the machine on every check. A project's .claude/council-heads-up.json overrides this for that project and picks the seat and cadence.",
  65. "default": false
  66. },
  67. "gemini_api_key": {
  68. "type": "string",
  69. "title": "Gemini API key",
  70. "description": "The key the gemini seat uses, kept in Claude Code's secure storage. The council's mod sets GEMINI_API_KEY from it for every command in the session, over a shell export.",
  71. "sensitive": true
  72. },
  73. "openai_api_key": {
  74. "type": "string",
  75. "title": "OpenAI API key",
  76. "description": "The key the openai seat uses, kept in Claude Code's secure storage. The council's mod sets OPENAI_API_KEY from it for every command in the session, over a shell export.",
  77. "sensitive": true
  78. },
  79. "xai_api_key": {
  80. "type": "string",
  81. "title": "xAI API key (Grok)",
  82. "description": "The key the grok seat uses, kept in Claude Code's secure storage. The council's mod sets XAI_API_KEY from it for every command in the session, over a shell export.",
  83. "sensitive": true
  84. },
  85. "perplexity_api_key": {
  86. "type": "string",
  87. "title": "Perplexity API key",
  88. "description": "The key the perplexity seat uses, kept in Claude Code's secure storage. The council's mod sets PERPLEXITY_API_KEY from it for every command in the session, over a shell export.",
  89. "sensitive": true
  90. },
  91. "kimi_api_key": {
  92. "type": "string",
  93. "title": "Kimi API key",
  94. "description": "The key the kimi seat uses, kept in Claude Code's secure storage. The council's mod sets KIMI_API_KEY from it for every command in the session, over a shell export.",
  95. "sensitive": true
  96. },
  97. "openrouter_api_key": {
  98. "type": "string",
  99. "title": "OpenRouter API key",
  100. "description": "The key every OpenRouter seat uses, kept in Claude Code's secure storage. The council's mod sets OPENROUTER_API_KEY from it for every command in the session, over a shell export.",
  101. "sensitive": true
  102. },
  103. "openrouter_models": {
  104. "type": "string",
  105. "title": "OpenRouter models",
  106. "description": "Model ids from openrouter.ai/models, separated by commas or spaces, such as deepseek/deepseek-v3.2,z-ai/glm-5.3. Each id becomes its own seat, openrouter-1, openrouter-2 and so on, all on the one OpenRouter key. The council's mod sets OPENROUTER_MODELS from it for every command in the session, over a shell export. Left empty, a shell export still applies; with neither, one openrouter seat runs the default model."
  107. }
  108. }
  109. }
README.md · README · 341 lines
  1. ![claude-council](assets/banner.svg)
  2. A Claude Code plugin that asks AI coding agents the same question in
  3. parallel. Compare their answers when one model's bias could mislead you:
  4. architecture decisions, debugging dead ends, security reviews, or framework
  5. choices.
  6. ![Codex, Antigravity, Gemini, and OpenAI answer the SQLite-or-Postgres question while fake Kimi fails](assets/council-run.svg)
  7. [![Tests](https://github.com/hex/claude-council/actions/workflows/tests.yml/badge.svg)](https://github.com/hex/claude-council/actions/workflows/tests.yml)
  8. ## Where it runs
  9. | Where | What you get |
  10. | --- | --- |
  11. | Claude Code in a terminal | `/claude-council:ask`, every seat, and the pane beside the chat. |
  12. | Claude Desktop, Code tab | The same plugin and the same commands. The pane opens in Desktop's side panel. |
  13. | Claude Desktop, an ordinary chat | The [Desktop extension](#claude-desktop): five tools, a seat picker and an answers panel in the chat. macOS and Linux. |
  14. ![Claude Desktop's Code tab: Claude's synthesis of a primary-key question on the left, the Council pane with each seat's answer on the right](assets/desktop-code-tab.png)
  15. In both, Claude can also call the council as a tool. It asks which seats and
  16. how verbose first, then shows a Send card, and nothing leaves your machine
  17. until you confirm it.
  18. claude.ai chat and Cowork run in a sandbox that can block calls to other
  19. providers, so the council does not work there.
  20. ## Install
  21. You need `curl` and `jq` on macOS, Linux, or Windows through Git Bash.
  22. ```text
  23. /plugin marketplace add hex/claude-marketplace
  24. /plugin install claude-council
  25. ```
  26. Give it at least one provider in the shell you start Claude Code from:
  27. ```bash
  28. export OPENAI_API_KEY="..."
  29. # Or GEMINI_API_KEY, XAI_API_KEY, PERPLEXITY_API_KEY,
  30. # KIMI_API_KEY, OPENROUTER_API_KEY
  31. ```
  32. Or store the keys in the plugin's config. Run `/plugin configure claude-council`,
  33. fill in the API key rows, and choose Save configuration. Claude Code keeps them in
  34. its secure storage, not in a settings file. The council's mod sets each stored
  35. key as its variable for the whole session, so every command Claude Code starts,
  36. in every project, gets the stored key in place of one your shell exports.
  37. The OpenRouter models row takes the model ids `OPENROUTER_MODELS` takes, so
  38. the router's seats need no export either.
  39. Stored keys need Claude Code 2.1.287 or later. The desktop app's plugin
  40. Settings tab takes them too, for sessions on that computer. They never reach
  41. claude.ai chat or Cowork. [API keys](docs/REFERENCE.md#api-keys) covers where they are kept,
  42. when the mod does not load, and how a cleared key behaves.
  43. The `codex`, `agy` (Antigravity), `grok`, `kimi`, and `cursor-agent` CLIs use
  44. your existing subscriptions without an API key. Claude Code itself, the
  45. Copilot CLI and OpenCode can sit too, when named in `--providers`. A local
  46. `ollama` model needs no key or external network. With none of these, `--local` uses Claude
  47. subagents. See [provider selection](docs/REFERENCE.md#choosing-providers) and
  48. [local councils](docs/USAGE.md#local-council---local).
  49. ## First run in 60 seconds
  50. Read [what gets sent and saved](docs/REFERENCE.md#what-gets-sent-and-saved)
  51. before asking about your code.
  52. ```text
  53. /claude-council:status
  54. /claude-council:ask "Should I use UUID or BIGINT primary keys for a SaaS users table?"
  55. ```
  56. Each provider answers independently. Claude writes a synthesis of their
  57. agreement, disagreements, and recommendation. Agreement is not verification:
  58. the models share your description and its assumptions.
  59. [The synthesis names the premise that could change the answer](docs/REFERENCE.md#reading-the-synthesis).
  60. ## Claude Desktop
  61. The plugin needs Claude Code. For an ordinary Claude Desktop chat there is a
  62. Desktop extension, `council-<version>.mcpb`, that gives Claude five tools:
  63. `council_ask`, `council_pick`, `council_result`, `council_status` and
  64. `council_settings`. It runs on macOS and Linux only, because it starts the
  65. council's bash scripts.
  66. 1. Download `council-<version>.mcpb` from the
  67. [latest release](https://github.com/hex/claude-council/releases/latest).
  68. To build it yourself, run `bash extension/build.sh` in a clean checkout.
  69. It needs `npm` and writes `dist/council-<version>.mcpb`.
  70. 2. Double-click the file. Desktop warns that the extension gets access to
  71. everything on your computer and is not verified by Anthropic. Both hold for
  72. any extension installed from a file: its server runs as you, with your files
  73. and network, and Anthropic has not reviewed it. Install only a package you
  74. built or trust.
  75. 3. In Settings, Extensions, Claude Council, Configure, paste a key for each
  76. provider you want: Gemini, OpenAI, xAI, Perplexity, Kimi, OpenRouter. An
  77. empty field leaves that provider's API out. Desktop stores the keys. For a
  78. question, each key goes only to its own vendor's seats: the vendor's API,
  79. and its CLI agent when one is installed (`codex` gets the OpenAI key, `agy`
  80. the Gemini key, `grok` the xAI key, `kimi` the Kimi key). A status check is
  81. different: it starts each installed CLI agent but `claude` and `opencode`
  82. with every key set.
  83. 4. In a chat, ask Claude to open the council settings. A panel in the chat
  84. lists each keyed provider's models, fetched live from that provider. Pick
  85. one per provider, or switch on up to 12 OpenRouter models, one seat each.
  86. Every change saves at once and applies to the next question.
  87. 5. Ask Claude to put a question to the council. Desktop asks you to approve
  88. the tool call, and that approval is the only check before the question
  89. leaves your machine. A panel under the tool call shows each seat's answer,
  90. or its error, in a card of its own headed by the seat's name and its
  91. vendor's logo, and Claude writes the synthesis below it. Desktop waits for
  92. the whole run, from seconds to a few minutes.
  93. ### Asking at once
  94. Name the seats or a verbosity (standard, brief or
  95. detailed) in your message, or say to just ask everyone, and Claude runs the
  96. question at once, on every default seat at standard unless you named others.
  97. Claude passes your question as you asked it, adding context only when it
  98. depends on earlier messages the council cannot see.
  99. ### The picker
  100. Otherwise, as in "ask the council: which queue should we
  101. use?", Claude opens `council_pick`, a panel with the question, a switch per
  102. seat and the three verbosities. Claude then waits, in the same turn, in
  103. `council_result` for the picker's run.
  104. 1. Press Ask. The answers appear in that panel, and the line under them says
  105. "Claude is writing the synthesis".
  106. 2. Claude writes the synthesis with no message from you. When every seat
  107. fails, Claude says what failed.
  108. When no seat answers, the panel says so and Ask works again.
  109. ### Changing a model in the picker
  110. On an API seat or an OpenRouter seat the
  111. model in grey is a button: click it to pick another model from that
  112. provider's live list (OpenRouter's has a filter box). The choice saves at
  113. once, as the settings panel's would, and applies to this question and the
  114. ones after it. An OpenRouter seat changes only its own model; adding or
  115. removing OpenRouter seats stays in the settings panel.
  116. ### When Claude is no longer waiting
  117. Claude waits at most 11.5 minutes from
  118. when the picker opened. If it has stopped waiting by the time the answers are
  119. in, or you press Ask a second time in the same panel:
  120. - Desktop puts "Synthesise the council's answers (run 3f9a2c)." in the
  121. message box, under a caution about malicious content, and the line under
  122. the answers says to press Enter. The panel cannot send that message itself.
  123. - Claude then fetches that run's answers with `council_result` and writes the
  124. synthesis.
  125. - If Desktop refuses the message, the line instead gives the words to send
  126. yourself (`Ask Claude: "synthesise council run 3f9a2c"`).
  127. - The extension keeps the last 5 runs in which a seat answered; after that,
  128. ask the council again.
  129. A question with no providers named goes to every provider with a key and to
  130. every CLI agent the scripts find (`codex`, `agy`, `grok`, `kimi`,
  131. `cursor-agent`, `ollama`) in your PATH, `~/.local/bin`, `~/bin`,
  132. `/opt/homebrew/bin` or `/usr/local/bin`. `claude`, `copilot` and `opencode`
  133. get it only when the question names them. A CLI agent answers on its own
  134. login, so it receives the question even when its vendor's key field is
  135. empty; apart from its own vendor's key it gets no key from the form. The
  136. [reference](docs/REFERENCE.md#claude-desktop-extension) lists the files it
  137. keeps and its limits.
  138. ## Features
  139. | Capability | Guide |
  140. |---|---|
  141. | Ask providers together, or name one in plain words | [Command examples](docs/REFERENCE.md#command-examples) |
  142. | Choose providers and pin a standing roster | [Provider selection](docs/REFERENCE.md#choosing-providers) |
  143. | Assign specialist roles | [Roles](docs/USAGE.md#specialized-roles) |
  144. | Have providers challenge each other's answers | [Debate](docs/USAGE.md#debate-mode) |
  145. | Gather code context with agents before asking | [Agent mode](docs/USAGE.md#agent-enhanced-analysis---agents) |
  146. | Use Claude subagents without an external provider | [Local council](docs/USAGE.md#local-council---local) |
  147. | Include relevant files or choose your own | [Auto-context](docs/USAGE.md#auto-context-injection) |
  148. | Attach a screenshot | [Image input](docs/USAGE.md#image-input) |
  149. | Ask about the current conversation | [`/advise`](docs/USAGE.md#asking-about-the-conversation-itself) |
  150. | Run in the background and fetch the result | [Background jobs](docs/USAGE.md#background-jobs---async) |
  151. | Export the answers to a file | [Export](docs/USAGE.md#export-to-file) |
  152. | Show only the synthesis | [Quiet mode](docs/USAGE.md#quiet-mode) |
  153. | Watch answers arrive and retry failed seats | [The pane](#the-pane) |
  154. | Get a note from one seat while Claude works | [Heads-up side-check](#heads-up-side-check) |
  155. | Have a second model review the diff before Claude stops | [Stop-gate review](docs/USAGE.md#stop-gate-review-opt-in-off-by-default) |
  156. | Ask the council from a Claude Desktop chat | [Claude Desktop](#claude-desktop) |
  157. | Call the scripts without Claude Code | [Direct script usage](docs/REFERENCE.md#direct-script-usage) |
  158. The [reference](docs/REFERENCE.md) lists every flag, variable, model default,
  159. and limit. [Architecture](docs/ARCHITECTURE.md) covers the implementation.
  160. [TESTING.md](TESTING.md) covers tests, releases, and adding a provider.
  161. ## Heads-up side-check
  162. ![The intro hint, a heads-up note, Learn more, and an unsent Tell Claude draft](assets/heads-up.svg)
  163. Opt in through the "Heads-up side-check" row in `/config` or the project's
  164. `.claude/council-heads-up.json`. One seat, codex by default, checks recent
  165. turns and tool results while Claude works. For a Write or an Edit, it gets the
  166. text Claude wrote. The note shows what went where.
  167. Learn more expands it; Tell Claude puts a draft in the prompt without sending
  168. it. Read the [data limits and off switch](docs/REFERENCE.md#what-gets-sent-and-saved)
  169. and the [heads-up guide](docs/USAGE.md#heads-up-side-check-opt-in-off-by-default).
  170. ## The pane
  171. On Claude Code 2.1.287 or later, the pane opens inside Claude Code,
  172. as in the animation above. Each seat has one row with its model, a bar for
  173. its state, its time, and a button for what it does now: cancel, retry, open,
  174. or fold. When the run ends,
  175. the synthesis opens at the top with the answers folded below it. Reopen the
  176. last run with `/council-pane`. Inside tmux, the mod asks once which pane you want.
  177. The [mod guide](mods/council-pane/README.md) lists keys, settings, limits,
  178. and when the mod does not load. It also lets Claude call the council as a tool;
  179. every call asks you first.
  180. ### What the mod runs, reads and sends
  181. The mod makes no network calls of its own. It starts the programs below
  182. from an argument array, never a shell string, and never hands `bash -c` a
  183. script. Your text reaches model providers only through the first two:
  184. - `bash scripts/run-council.sh` when Claude calls the council tool and you
  185. choose `Send to the council`. It sends your question to the providers you
  186. set up.
  187. - `bash scripts/heads-up.sh`, only with the heads-up side-check on. It reads
  188. this session's transcript (by default the last 5 turns and up to 8 tool
  189. results of 2,000 characters each) and sends that digest to one seat,
  190. `codex` unless you pick another. The run directory keeps `digest.md` and
  191. `prompt.md`, exactly what left the machine, until the mod reads the result
  192. and removes it. The codex seat runs in that directory under a codex home
  193. of its own: your model and effort, a link to your `auth.json` so a token
  194. refresh reaches your login (without one, your `CODEX_API_KEY` or
  195. `OPENAI_API_KEY` signs codex in), and none of your plugins, hooks,
  196. `AGENTS.md`, skills or MCP servers.
  197. - `bash scripts/query-council.sh --list-default-models` when Claude calls
  198. the council tool, to list the seats it may offer you and to check the
  199. seats it names. It reads your configuration and sends nothing.
  200. - `bash scripts/seat-info.sh discover`, `model <seat>` or
  201. `allowlisted <seat>` during `/heads-up setup` and `/heads-up on <seat>`.
  202. It checks which seats this machine can run and sends nothing.
  203. - `codex debug models` during `/heads-up setup` with the codex seat, to list
  204. its models. That command belongs to codex, and so does any request it
  205. makes.
  206. - `kill -0 <pid>` to check a council run is still alive. It sends no
  207. signal.
  208. - `rm -rf` and `mv -f` on its own run directories under
  209. `$TMPDIR/council-mod.<session id>`, to clean up a finished run and to
  210. answer the retry offer.
  211. It writes files in that `$TMPDIR` directory, including `model.json` and
  212. `avoid.txt`, which `heads-up.sh` reads. Outside it, it writes one file,
  213. `.claude/council-heads-up.json` in the project, and only from
  214. `/heads-up setup`, `/heads-up on`, `/heads-up off` or a note's `Turn off`.
  215. That file is the heads-up switch. Claude Code itself does not read it.
  216. It sets environment variables in Claude Code's own process. With an API key
  217. or the OpenRouter models list stored in the plugin's config, it sets that
  218. variable, such as `OPENAI_API_KEY` or `OPENROUTER_MODELS`, when the session
  219. starts. Every command Claude Code starts after that gets the stored value in
  220. place of the one your shell exports, the same as an export would. Before a council run it sets `COUNCIL_MOD_PANE_DIR`.
  221. It touches the prompt in two places. With `wake_on_async_done` on, it
  222. submits `The background council job <id> has finished. Fetch it with
  223. /claude-council:result <id> and summarise it.` once a background job
  224. finishes. A note's `Tell Claude` puts a draft in the prompt box, and you decide
  225. whether to send it.
  226. It reads the text of each Bash command Claude runs, only to spot a call
  227. to `run-council.sh`. Just before that call it sets `COUNCIL_MOD_PANE_DIR`,
  228. so the run draws in this pane, and clears it when tmux owns the pane. It reads `TMPDIR` and `TMUX`. It turns on the
  229. `claude-council.heads_up` setting only when you pick "every project" in
  230. `/heads-up setup`.
  231. ![Real SQLite-or-Postgres answers arrive in tmux beside a fake Kimi failure and the retry offer](assets/tmux-pane.svg)
  232. Without the mod, tmux opens a side pane that streams answers and each
  233. provider's status. It offers to retry failed seats. Esc closes it.
  234. `--no-pane` or `COUNCIL_NO_PANE=1` turns it off. Outside both pane modes,
  235. answers arrive in the chat. See [terminal integration](docs/REFERENCE.md#display--terminal-integration).
  236. The council and tmux animations show real answers to
  237. "Should a single-node demo app use SQLite or Postgres?" Codex and Antigravity
  238. use subscriptions; Gemini and OpenAI use their APIs. Kimi is a rate-limit
  239. fixture. The heads-up animation uses a fake seat and a real Claude Code host.
  240. [Capture scripts and raw frames](assets/capture/README.md) record how to
  241. reproduce them.
  242. ## What a CLI seat loads
  243. Every CLI seat runs from an empty directory, `$TMPDIR/claude-council-<uid>/seat`,
  244. where `<uid>` is your numeric user id. The `CLAUDE.md`, `AGENTS.md`, rules,
  245. skills, hooks and MCP servers of the project you ask from don't reach it. The
  246. directory doesn't limit what a seat can read: codex, for example, can still
  247. open any file by its full path.
  248. The directory is read-only. That stops a seat writing a file there, but not a
  249. seat that changes the mode first. A seat also refuses to start when it finds
  250. anything in it, and names the path so you can look before you remove it. A file
  251. that lands while other seats are running can reach them; the next seat to start
  252. refuses. A seat also refuses to start when `TMPDIR` sits inside a git
  253. repository, and names it: a CLI would take the whole repository as the seat's
  254. project.
  255. The seats below also run under a home of their own, in
  256. `$TMPDIR/claude-council-<uid>/seat-home`. It links only to what the CLI needs to
  257. sign in and start. Only you can open either directory, and a seat refuses to
  258. start when one of them belongs to another user or is a symlink.
  259. | Seat | No longer loads | Still loads |
  260. |---|---|---|
  261. | `cursor-cli` | Skills and plugins from `~/.cursor`, `~/.claude`, `~/.codex`, `~/.grok` and `~/.agents`, Cursor and Claude hooks, `~/.cursor/mcp.json` | Team commands from your Cursor dashboard (read from the CLI's bundle, not seen in a run) |
  262. | `grok-cli` | Claude plugins, `~/.agents` skills, grok's memory across sessions | Your `~/.grok` config and the hooks in `~/.grok/hooks`, since no grok setting turns them off. The seat home links your login keychain (macOS) and `~/.docker/run`, for Docker Desktop's socket |
  263. | `antigravity` | On macOS: your `~/.gemini` rules, skills, MCP servers and saved conversations | On Linux: everything under `~/.gemini`, because where agy keeps its login there is not confirmed |
  264. | `opencode` | The `~/.claude` and `~/.agents` skills, agents and `CLAUDE.md` | Your login and session history in `~/.local/share/opencode` |
  265. | `codex` | Your plugins, hooks, skills, memories and MCP servers. Each question gets its own codex home with your model, your effort and a link to `auth.json` | Nothing. With no `auth.json` and `cli_auth_credentials_store` set to `keyring` or `auto`, your login is in the system keyring, which codex can't read from another home, so the seat runs under your own codex home and the answer starts with a note. An exported `OPENAI_API_KEY` doesn't change that; an exported `CODEX_API_KEY` still signs codex in ahead of any login, as it does when you run codex yourself. Set `cli_auth_credentials_store = "file"` and run `codex login` again to isolate it. With no login at all, the seat stops and says so |
  266. | `claude-cli` | The git status and commit instructions a repository adds | Admin policy and your organisation's instructions |
  267. | `kimi-cli`, `copilot-cli` | The project's files | Your home: kimi's hooks in `~/.kimi-code/config.toml`, copilot's skills in `~/.copilot` and `~/.agents` |
  268. Cursor and Antigravity keep the model you picked in their own config: the seat
  269. reads it from your real home and passes it on. On Windows the seats keep your
  270. home, because the native CLIs find it through `USERPROFILE`.
  271. [Seat isolation](docs/ARCHITECTURE.md) has each seat's details.
  272. ## Troubleshooting
  273. | Problem | Fix |
  274. |---|---|
  275. | Cloned the repo and nothing loads | Don't clone into `~/.claude/plugins/`. `pluginDirectories` is not a setting. Follow [running from a local clone](docs/REFERENCE.md#running-from-a-local-clone). |
  276. | No providers found | Run `/claude-council:status`. Export keys in the shell that starts Claude Code, store them with `/plugin configure claude-council`, or use `--local`. |
  277. | A stored key is not used | After saving one with `claude plugin configure` in a shell, restart Claude Code. Stored keys need the council's mod, which loads on Claude Code 2.1.287 or later and not under `--safe-mode` or `disableAllHooks`. Without it, export the key in your shell. |
  278. | A different model answered | A header such as `grok-4.7 (grok-latest unavailable)` records a [model fallback](docs/REFERENCE.md#model-fallback). Setting that provider's `_MODEL` variable disables fallback. |
  279. | No pane | Use tmux, or update to Claude Code 2.1.287 or later for the pane inside Claude Code. The [mod guide](mods/council-pane/README.md) lists what stops it loading. |
  280. | Updating | Run `/plugin update claude-council`, then restart Claude Code. |

Version history

  • v2026.11.211/18/2026

  • v2026.11.111/9/2026

  • v2026.10.1410/27/2026

  • v2026.10.1310/26/2026

  • v2026.10.1110/24/2026

  • v2026.10.1010/23/2026

  • v2026.10.710/21/2026

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.