LLM Council

Karpathy's LLM Council pattern with Anthropic and OpenAI-compatible providers.

When to use

• High-stakes plan review (/plan crosses N-file threshold)
• Conflicting learning-rules → re-resolve via vote
• User invokes /council "<query>" or /wiki council
• Architecture decisions where you want multiple viewpoints captured
• Persisting deliberation as a wiki page for future reference

Three phases

1. Independent: each model answers in parallel
2. Ranking: each model ranks anonymized peer responses
3. Synthesis: chairman model reads all responses + rankings → final answer

Provider config

Configure optional keys in the plugin configuration dialog. Use the providers MCP server's run_provider_task tool with task: "council" and args: ["run", "<query>", "--provider", "openai"]. For provider status, pass args: ["providers"]. Never request keys in chat or read them from existing credentials or shell configuration.

For standalone CLI use, explicitly set a pro-workflow-specific variable below. The first configured provider is the default; --provider selects one explicitly. See provider configuration.

Override per-run with --provider openai|anthropic|openrouter|fireworks|custom.

Default model rosters per provider live in scripts/council.js and can be overridden via --models CSV and --chairman <id>.

Commands

In a plugin session, pass these runner arguments through run_provider_task. The direct commands below are for standalone CLI installations with explicit credentials.

node $SKILL_ROOT/scripts/council.js run "<query>" [--models id1,id2,id3] [--chairman id] [--provider <name>] [--wiki <slug>]
node $SKILL_ROOT/scripts/council.js providers
node $SKILL_ROOT/scripts/council.js show <session-id>

--wiki <slug> writes the full transcript to <wiki>/derived/council/<session-id>.md and registers it via wiki-cli.js page so it shows in FTS5 search.

Output

Each session writes:

~/.pro-workflow/council/<session-id>/
├── config.json           # query, models, chairman, provider
├── phase1_responses.json # raw API responses per model
├── phase2_rankings.json  # anonymized ranking outputs
├── phase3_synthesis.txt  # chairman's final answer
└── final_output.md       # human-readable bundle

Console prints the markdown bundle. Pipe to pbcopy / tee as needed.

Hard rules

1. Never skip the ranking phase. It's the core of the council pattern.
2. Save raw responses to disk verbatim. No summarization in storage.
3. Anonymize responses for ranking — models see Response A/B/C/..., not peer names.
4. The chairman sees both real names AND rankings.
5. Display all three phases to the user. No phase elision.

Cost awareness

The script logs per-call latency + tokens on supported providers. Multiply by your provider rate to estimate. Council cost grows linearly with len(models)^2 (each model ranks all others) plus the chairman.

Default council size: 3-5 models. More models = exponentially more ranking calls.

Use with wiki

/wiki council agent-memory "should we adopt episodic memory in our agents?"

Loads agent-memory wiki context as system prompt prefix, runs council, persists transcript as wiki/derived/council/<id>.md. The transcript becomes searchable via /wiki ask.