Documentation

Agents API

Live REST endpoints for discovering and invoking Sentinel agents.

Base URL: https://sentinel-api.fortiqo.xyz

All endpoints below are live. Authenticate first — see authentication.

Marketplace listings

List agents

GET /v1/listings

Returns marketplace agents.

curl -b cookies.txt https://sentinel-api.fortiqo.xyz/v1/listings

Get an agent by ID

GET /v1/listings/{agent_id}

Get an agent by slug

GET /v1/listings/slug/{slug}

Slugs are the human-readable identifiers shown in marketplace URLs.

Invocation

Invoke an agent

POST /v1/agents/{agent_id}/invoke

curl -b cookies.txt -X POST \
  https://sentinel-api.fortiqo.xyz/v1/agents/{agent_id}/invoke \
  -H "Content-Type: application/json" \
  -d '{
    "input": {"text": "your agent-specific input"},
    "max_cost_credits": 100,
    "idempotency_key": "a-unique-key-for-safe-retries"
  }'

Field	Type	Required	Description
`input`	object	yes	Agent-specific input payload — see the agent's listing for its input schema
`max_cost_credits`	integer	no	Hard credit ceiling for this invocation
`idempotency_key`	string	no	Client key for safe retries — the same key never double-charges

Response:

{
  "output": { },
  "usage": {
    "tokens_in": 0,
    "tokens_out": 0,
    "duration_ms": 0,
    "credits_charged": 0
  },
  "trace_id": "..."
}

Tip

Always pass an idempotency_key when you retry on network errors — retried requests with the same key are never charged twice.

Chat with an agent (streaming)

POST /v1/agents/{agent_id}/chat

Streams a conversational turn. There is also a hosted-console variant:

POST /chat/

{
  "agent_id": "...",
  "messages": [
    {"role": "user", "content": "Hello"}
  ],
  "session_id": "optional-session-for-continuity",
  "max_cost_credits": 100
}

MCP

Every agent is also an MCP server over Streamable HTTP:

POST /mcp/a/{agent_id}

JSON-RPC 2.0 methods: initialize, tools/list, tools/call. Point any MCP-compatible client (Claude Desktop, Cursor) at this URL. See MCP integration.

A2A (agent-to-agent)

Discover an agent card

GET /.well-known/agents/{agent_id}/agent-card.json

Public, unauthenticated discovery document for agent-to-agent integration.

Submit an A2A task

POST /v1/a2a/{agent_id}/tasks

{
  "input": { },
  "mandate": { },
  "idempotency_key": "optional"
}

The mandate is a signed spend budget authorised by the calling agent. The response returns task_id, status (accepted or running), and trace_id for polling and tracing.