MCP Integration

Vāda MCP

Vāda exposes deliberation as MCP tools (vada__consult and vada__deliberate). Any MCP-compatible client can call them. The hosted server runs at https://vada.attalabs.dev/api/mcp — no local install required. Authentication via Vāda API key. Provider model calls run on your configured BYOK keys.

Getting started

Connect to the hosted server — no local setup required.

Add Vāda as a custom MCP connector with the server URL and an Authorization header carrying your Vāda API key.

{
  "url": "https://vada.attalabs.dev/api/mcp",
  "headers": {
    "Authorization": "Bearer vada_your_key_here"
  }
}

API Key

Generate a Vāda API key in Settings → Vāda API Keys. It is shown once — copy it then. Treat it like a password— it gives full access to your account's deliberation tools.

Generate a key

Go to Settings → Vāda API Keys and generate a new API key.

Add it to your client

Paste the key as the Authorization: Bearer value in your client config above.

Provider Keys

The hosted server runs on your provider keys. Configure them in Settings → Provider API Keys. Keys are encrypted at rest. Over MCP, Vāda uses your Anthropic, Google, OpenAI and xAI keys; an Anthropic key is needed for the default models.

Vāda's server decrypts your keys to make provider calls on your behalf, and stores each consultation in full. See the trust page for more detail.

Tools

vada__consult

Independent reviewers in parallel. Pass your context, the question and 2–5 reviewers (strategist, critic, devils_advocate). Returns each reviewer's response and a link to the full session. Pass spec_id: "vada-fusion-native" for an Outside Read.

Use vada__consult to get an outside read on this plan: [your question and context]
Reviewers: strategist, critic, devils_advocate

vada__deliberate

Run a team on your question. Pass a question and optionally a team. Returns the team's output, its structured result when the team defines one, and a link to the full session. Default team: vada-reviewers.

Use vada__deliberate to analyze: [your question here]
Team: vada-reviewers-synthesis

Teams

The teams each tool accepts.

vada-reviewers Reviewers · vada__deliberate

Three reviewer slots run in parallel. You choose the model for each slot — any vendor you have a key for. Each reviewer has live web access.

vada-reviewers-synthesis Reviewers + Synthesis · vada__deliberate

The same three slots, then a synthesizer extracts consensus, contradictions and unique insights into a structured analysis.

vada-fusion-native Outside Read · vada__consult (spec_id)

Four reviewers from different vendors examine your question in isolation. A synthesizer maps where they agree, what they concede and what stays contested; a blind critic and a fact-checker audit it before you get it.

Alternative

Local stdio server

For local development or self-hosted setups, run the MCP server as a local process on your own keys: clone the repo and point your client at the server's entry file.

Prerequisites

Bun (bun.sh)
An Anthropic API key (required); Google, OpenAI and xAI keys are optional
Clone the repo: git clone https://github.com/atta-labs/attalabs
Install dependencies: cd attalabs && bun install

Config file location

~/Library/Application Support/Claude/claude_desktop_config.json

Add to config

{
  "mcpServers": {
    "vada": {
      "command": "bun",
      "args": [
        "run",
        "/path/to/attalabs/apps/vada-ai/mcp-server/src/index.ts"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-...",
        "GOOGLE_API_KEY": "optional",
        "OPENAI_API_KEY": "optional",
        "XAI_API_KEY": "optional",
        "DATABASE_URL": "optional — postgresql://...",
        "VADA_USER_ID": "optional — your Clerk user id"
      }
    }
  }
}

DATABASE_URL and VADA_USER_ID are optional: set them to log consultations to your own database.

Restart

Quit Claude Desktop completely and relaunch. The MCP server starts automatically on next launch.

Verify

Open your client's tools or connectors menu — vada__consult and vada__deliberate should be listed.

Troubleshooting

Tool not appearing in client

Check that your API key is valid and that your client can reach vada.attalabs.dev. Some clients require a restart after config changes.

401 Unauthorized

The API key is invalid, revoked or not sent as Bearer vada_…. Generate a new one in Settings → Vāda API Keys.

400 "No provider keys configured"

The hosted server needs a provider key to make model calls. Configure one in Settings → Provider API Keys.

429 "Rate limit exceeded"

Each Vāda API key allows 60 requests per hour. Wait, or spread calls over more time.

Request times out

Calls typically take 30–120 seconds, longer for larger teams. Raise your client tool timeout to at least two minutes.