Roundtable

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Multi-model AI debates: GPT-4o, Claude, Gemini & 200+ models discuss, then synthesize insight.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "Roundtable": {
      "url": "https://mcp.roundtable.now/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (13)

• list-models read-only

  List available AI models grouped by thinking level (low/medium/high). Shows default models, credit costs, capabilities for each tier. Use this before consult to understand model options.

  Parameters:

  • thinking_level — string — Filter to a specific thinking level

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• list-sessions read-only

  List your previous MCP tool sessions. Returns session metadata including prompt, tool used, quality score, and credits consumed. Useful for reviewing past council discussions.

  Parameters:

  • limit — integer — Max results to return
  • offset — integer — Pagination offset
  • tool_name — string — Filter by tool name (e.g., "consult", "architect")

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get-session read-only

  Get full details of a previous MCP session by ID. Returns the complete result including participant responses and moderator synthesis. Use list-sessions first to find session IDs.

  Parameters:

  • session_id (required) — string — The session ID to retrieve

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get-logs read-only

  Query structured logs from your MCP tool executions. Filter by session, severity level, event type, and time range. Useful for debugging and monitoring tool usage.

  Parameters:

  • event — string — Filter by event name (e.g., "debate_completed")
  • level — string — Filter by log level
  • limit — integer — Max results
  • offset — integer — Pagination offset
  • end_time — number — End timestamp (ms) for time range filter
  • session_id — string — Filter logs for a specific session
  • start_time — number — Start timestamp (ms) for time range filter

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• check-usage read-only

  Check your remaining credits, usage limits, and plan info

  Parameters:

  • verbose — boolean — Include detailed per-window rate limit breakdown

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get-thread-link read-only

  Get the dashboard URL for a previous debate session. Returns the thread link and public URL if the thread is public.

  Parameters:

  • session_id (required) — string — The session ID to get the thread link for

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• set-thread-visibility

  Set a thread as public or private. Public threads can be shared via URL. Use session_id from a previous debate session.

  Parameters:

  • is_public (required) — boolean — Set to true to make the thread publicly accessible, false to make it private
  • session_id (required) — string — The session ID from a previous debate

  Hints: read-only no · destructive no · idempotent yes · open-world no

• consult-council

  Consult the AI coding council — multiple models discuss your engineering question sequentially (each sees prior responses), then a moderator synthesizes. Auto-mode by default — AI picks optimal models, roles, and conversation mode from your prompt. Provide explicit models to override (manual mode). Fully configurable: mode, format, roles, models, thinking level.

  Parameters:

  • prompt (required) — string — The question, topic, or problem to debate
  • mode — string — Conversation mode: analyzing (research), brainstorming (ideas), debating (tradeoffs), solving (action plans)
  • roles — array — Inline role names for participants (e.g., ["Security Architect", "Backend Engineer"])
  • format — string — Moderator output format: discussion (narrative), adr (architecture decision), comparison (table), pros-cons
  • models — array — Override specific model IDs. Min 3 models. Use list-models to see available options
  • context — string — Additional background context for the debate (code, docs, requirements)
  • knowledge — array — Reference knowledge to inject as context
  • auto_route — boolean — Auto-select optimal models based on prompt analysis and historical performance
  • webhook_url — string — Webhook URL to POST results to after completion
  • thinking_level — string — Controls model quality and cost: low (fast/cheap), medium (balanced), high (maximum reasoning)
  • session_context — array — Session IDs to use as context (max 3). Prior moderator summaries will be prepended.

  Hints: read-only no · destructive no · idempotent no · open-world yes

• design-architecture

  Architecture design council. Systems Architect, Infrastructure Engineer, and DX Advocate evaluate your system design. Always uses high thinking for maximum depth. Output as ADR.

  Parameters:

  • description (required) — string — What the system should do
  • scale — string — Target scale: startup (small team), growth (scaling), enterprise (large org)
  • tech_stack — array — Preferred technologies
  • focus_areas — array — Priority areas (e.g., ["security", "performance"])
  • webhook_url — string — Webhook URL to POST results to after completion

  Hints: read-only no · destructive no · idempotent no · open-world yes

• review-code

  Code review council. Senior Engineer, Security Reviewer, and Performance Analyst analyze your code and a moderator synthesizes their findings.

  Parameters:

  • code (required) — string — The code to review
  • focus — array — Review focus areas (e.g., ["security", "performance"])
  • language — string — Programming language (auto-detected if not specified)
  • webhook_url — string — Webhook URL to POST results to after completion
  • thinking_level — string — Review depth: low (quick scan), medium (balanced), high (thorough)

  Hints: read-only no · destructive no · idempotent no · open-world yes

• plan-implementation

  Implementation planning council. Tech Lead, Senior Engineer, and QA Strategist break down a feature into actionable steps, identify risks, and define acceptance criteria. Output as ADR.

  Parameters:

  • feature (required) — string — The feature or change to plan
  • knowledge — array — Reference knowledge to inject as context
  • tech_stack — array — Current tech stack
  • constraints — array — Constraints (e.g., ["no breaking changes", "must support offline"])
  • webhook_url — string — Webhook URL to POST results to after completion
  • thinking_level — string — Planning depth
  • session_context — array — Session IDs to use as context (max 3). Prior moderator summaries will be prepended.
  • codebase_context — string — Relevant existing code, file structure, or architecture notes

  Hints: read-only no · destructive no · idempotent no · open-world yes

• debug-issue

  Debugging council. Root Cause Analyst, Systems Engineer, and Edge Case Investigator collaboratively diagnose bugs, analyze errors, and propose fixes.

  Parameters:

  • problem (required) — string — Describe the bug, failure, or unexpected behavior
  • code — string — The relevant code where the bug occurs
  • error — string — Error message, stack trace, or unexpected output
  • knowledge — array — Reference knowledge to inject as context
  • webhook_url — string — Webhook URL to POST results to after completion
  • thinking_level — string — Analysis depth
  • session_context — array — Session IDs to use as context (max 3). Prior moderator summaries will be prepended.
  • expected_behavior — string — What should happen vs what actually happens

  Hints: read-only no · destructive no · idempotent no · open-world yes

• assess-tradeoffs

  Tradeoff assessment council. Pragmatist, Skeptic, and Futurist evaluate options from different angles — short-term vs long-term, risk vs reward, simplicity vs flexibility. Output as pros-cons.

  Parameters:

  • decision (required) — string — The decision or question to evaluate
  • context — string — Background context — codebase, team, timeline, constraints
  • options — array — Specific options to compare
  • priorities — array — What matters most (e.g., ["performance", "dx", "cost"])
  • webhook_url — string — Webhook URL to POST results to after completion
  • thinking_level — string — Analysis depth
  • session_context — array — Session IDs to use as context (max 3). Prior moderator summaries will be prepended.

  Hints: read-only no · destructive no · idempotent no · open-world yes

Repository

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.