Quick Start

Prerequisites

• A FleetRun account — sign up at fleetrun.app
• An API key from Anthropic, OpenAI, or an OpenRouter-compatible provider

Step 1: Connect a provider

After signing in, go to Settings → Providers and paste your API key. FleetRun supports:

• Anthropic — Claude 3.5, Claude 3.7 Sonnet, Claude Opus 4
• OpenAI — GPT-4o, GPT-4.1, o3, o4-mini
• OpenRouter — 200+ models via a single key

Keys are encrypted at rest using AES-256-GCM and never logged.

Step 2: Create an agent

Navigate to Agents and click New agent. Fill in:

• Name — human-readable identifier (e.g. Researcher)
• Role — system prompt describing what the agent does
• Model — pick from your connected providers
• Tools (MCPs) — optional MCP servers the agent can call

Example role prompt for a research agent:

You are a research specialist. When given a topic:
1. Break it into sub-questions.
2. Answer each sub-question with specifics and sources.
3. Synthesize a final summary with key takeaways.
Be concise and factual. Avoid speculation.

Step 3: Create and run a task

Go to Tasks and click New task. Fill in the task title and input, then assign it to your agent. Click Run to start execution. You will see a real-time stream of:

• Agent thinking steps
• Tool calls (if MCPs are attached)
• Intermediate messages
• Final output

Via the API

You can also create and trigger tasks programmatically. First, get your API token from Settings → API Tokens. Then:

# Create a task
curl -X POST https://app.fleetrun.app/api/tasks \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Research: state of AI agents in 2026",
    "input": "Summarize the current landscape of autonomous AI agents.",
    "assignee": "researcher",
    "priority": "normal"
  }'

# Run the task (returns SSE stream)
curl -X POST https://app.fleetrun.app/api/run \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{ "taskId": "T-1234" }'

Step 4: Observe output

Every task run is logged to the message bus. You can view the full execution trace in the task dock under the Logs tab. Logs include timestamps, model used, token counts, and any tool call inputs/outputs.

Environment variables

If you are self-hosting FleetRun, the following environment variables are required:

# Database
DATABASE_URL=postgresql://...

# Redis (for BullMQ task queue)
REDIS_URL=redis://...

# Auth
SESSION_SECRET=<random-32-byte-hex>

# LLM providers (at least one required)
ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
OPENROUTER_API_KEY=sk-or-...

# Telegram bot (optional — for agent notifications)
TELEGRAM_BOT_TOKEN=...