Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Councly MCP Server create a council hearing to review this code for security vulnerabilities"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP (Model Context Protocol) server for Councly - Multi-LLM Council Hearings.
Enable Claude Code, Codex, and other MCP-compatible AI assistants to invoke council hearings where multiple LLMs (Claude, GPT, Gemini, Grok) debate topics and synthesize verdicts.
npm install -g @councly/mcpOr use directly with npx:
npx @councly/mcpRelated MCP server: trinity-local
Sign in to Councly
Go to Settings > MCP Integration
Create a new API key
Copy the key (shown only once)
Add to your Claude Code settings (~/.claude/settings.json):
{
"mcpServers": {
"councly": {
"command": "npx",
"args": ["@councly/mcp"],
"env": {
"COUNCLY_API_KEY": "cnc_your_key_here"
}
}
}
}export COUNCLY_API_KEY=cnc_your_key_hereOr add to your shell profile.
Create a council hearing where multiple LLMs debate a topic.
Parameters:
Parameter | Type | Required | Default | Description |
subject | string | Yes | - | The topic to discuss (10-10000 chars) |
preset | string | No | balanced | Model preset: |
workflow | string | No | auto | Workflow: |
wait | boolean | No | true | Wait for completion |
timeout_seconds | number | No | 300 | Max wait time (30-600) |
Presets:
Preset | Credits | Counsels | Best For |
balanced | 9 | 3 | General purpose discussions |
fast | 6 | 3 | Quick responses, simple topics |
coding | 14 | 3 | Code review, technical decisions |
coding_plus | 17 | 4 | Complex code problems |
Example:
Use councly_hearing with subject="Review this Python function for security issues:
def authenticate(username, password):
query = f'SELECT * FROM users WHERE username='{username}' AND password='{password}''
return db.execute(query)
" and preset="coding"Check the status of a hearing.
Parameters:
Parameter | Type | Required | Description |
hearing_id | string (uuid) | Yes | The hearing ID |
Example:
Use councly_status with hearing_id="550e8400-e29b-41d4-a716-446655440000"Completed hearings return:
Status: completed, failed, or early_stopped
Verdict: Synthesized conclusion from the moderator
Trust Score: 0-100 confidence rating
Counsel Perspectives: Summary from each counsel
Cost: Credits used
Common errors:
Code | Description |
INSUFFICIENT_BALANCE | Not enough credits |
ACTIVE_HEARING_EXISTS | One hearing already in progress |
RATE_LIMIT_EXCEEDED | Too many requests |
CONTENT_BLOCKED | Subject contains prohibited content |
Variable | Required | Description |
COUNCLY_API_KEY | Yes | Your MCP API key |
COUNCLY_BASE_URL | No | API base URL (default: https://councly.ai) |
Councly uses a credit-based pricing model:
1 credit = $0.01 USD
Credits are deducted at hearing creation
Failed hearings are refunded
Purchase credits at councly.ai/billing.
Apache 2.0 - See LICENSE
Create a council hearing where multiple LLMs (Claude, GPT, Gemini, Grok) debate a topic and a moderator synthesizes the verdict.
Use cases:
Code review: Get diverse perspectives on code quality, architecture, security
Technical decisions: Compare approaches, weigh trade-offs
Problem solving: Generate and evaluate multiple solutions
Brainstorming: Explore ideas from different angles
The hearing runs asynchronously. By default, this tool waits for completion and returns the verdict. Set wait=false to get the hearing ID immediately and check status later with councly_status.
Cost: Varies by preset (6-17 credits). Check councly.ai for current pricing.
Inputs: subject (required), preset, workflow, wait, +1 more.
Check the status of a council hearing.
Returns:
For in-progress hearings: current phase and progress percentage
For completed hearings: verdict, trust score, and counsel summaries
For failed hearings: error message
Use this to check on hearings created with wait=false, or to retrieve past hearing results.
Inputs: hearing_id (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 2 tools
The two tools have clearly distinct purposes: councly_hearing creates and runs a council hearing, while councly_status checks the status or retrieves results of an existing hearing. There is no overlap in functionality, making it easy for an agent to select the correct tool based on the desired action.
Both tools follow a consistent naming pattern with the prefix 'councly_' followed by a descriptive action (hearing, status). This verb_noun style is uniform and predictable, aiding in tool identification and usage without confusion.
With only 2 tools, the server feels thin for its apparent scope of facilitating multi-LLM debates and status tracking. While the tools cover creation and status checking, the domain suggests potential gaps (e.g., no tool for listing hearings, modifying settings, or handling errors beyond status retrieval), making the count insufficient for comprehensive coverage.
The tool surface is significantly incomplete for the server's purpose. It lacks operations such as listing past hearings, canceling or deleting hearings, configuring hearing parameters beyond defaults, or managing user settings. This forces agents into dead ends for common workflows, like reviewing multiple past hearings or adjusting debate parameters.
Convene a panel of expert AI personas to debate any decision from every side.
Put a decision to a council of top AI models and get one brief: verdict, splits, risks, next steps.
Multi-LLM council: 25+ frontier models in parallel, consensus scoring, verdict-first code review.
Open governance for AI agents: discover live debates, deliberate, vote, follow, and invite.
