AI Agents

AI agents in inboard respond when @mentioned in workspace threads. They can answer questions, analyze data, and assist with financial modeling.

Setup

Step 1: Configure the API key

Set the CONCOURSE_AGENTS_API_KEY environment variable to your Anthropic API key:

docker run -d --name concourse \
  -p 8080:8080 \
  -v concourse-data:/data \
  -e CONCOURSE_JWT_SECRET=your-secret-at-least-32-bytes-long \
  -e CONCOURSE_ADMIN_EMAIL=admin@example.com \
  -e CONCOURSE_ADMIN_PASSWORD=changeme \
  -e CONCOURSE_AGENTS_API_KEY=sk-ant-xxx \
  ghcr.io/inboard-ai/concourse:latest

If no API key is configured, agent @mentions are silently ignored.

Step 2: Create an agent

Create an agent via the API:

curl -X POST http://localhost:8080/api/v1/agents \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Analyst",
    "model": "claude-sonnet-4-5-20250514",
    "system_prompt": "You are a financial analyst. Provide concise, data-driven answers."
  }'

Response:

{
  "id": "agent-uuid",
  "name": "Analyst",
  "provider": "anthropic",
  "model": "claude-sonnet-4-5-20250514",
  "system_prompt": "You are a financial analyst. Provide concise, data-driven answers.",
  "enabled": true,
  "created_by": "admin-uuid",
  "created_at": "2026-01-15T10:00:00Z"
}

Step 3: Use in workspace threads

Users @mention the agent by name in any workspace thread. The agent processes the message and posts a reply in the same thread.

Managing agents

List all agents:

curl http://localhost:8080/api/v1/agents \
  -H "Authorization: Bearer $TOKEN"

Update an agent:

curl -X PUT http://localhost:8080/api/v1/agents/{id} \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Senior Analyst",
    "model": "claude-sonnet-4-5-20250514",
    "system_prompt": "You are a senior financial analyst with expertise in M&A."
  }'

Delete an agent:

curl -X DELETE http://localhost:8080/api/v1/agents/{id} \
  -H "Authorization: Bearer $TOKEN"

Supported models

Agents use Anthropic Claude models. Supported model identifiers:

Model	Description
claude-sonnet-4-5-20250514	Balanced performance and speed.
claude-opus-4-0-20250514	Highest capability.
claude-haiku-3-5-20241022	Fastest responses, lower cost.

Set the model when creating or updating an agent.

Audit logging

Each agent invocation is logged as an audit event with:

• Token usage (input and output tokens)
• Latency
• The workspace and thread context

See Audit & Compliance for details.

API reference

See Helm (AI Agents) API for the complete endpoint reference.