Agent Gateway

A FastAPI extension for building API-first AI agent services. Define agents, tools, and skills as markdown files, then serve them as a production-ready API with authentication, persistence, scheduling, notifications, and more.

Quick Start

pip install agents-gateway[all]

# Scaffold a new project
agents-gateway init myproject
cd myproject

# Start the server
agents-gateway serve

Your agent API is now running at http://localhost:8000 with interactive docs at /docs.

Define an Agent

Create a markdown file at workspace/agents/assistant/AGENT.md:

---
description: A helpful assistant that answers questions
skills:
  - general-tools
memory:
  enabled: true
---

You are a helpful assistant. Answer questions clearly and concisely.

That's it — the agent is now available via the API.

Add a Tool

File-based tool

Create workspace/tools/http-example/TOOL.md:

---
name: http-example
description: Make an HTTP GET request and return the response
parameters:
  url:
    type: string
    description: The URL to fetch
    required: true
---

Add a handler in workspace/tools/http-example/handler.py:

import httpx

async def handler(url: str) -> str:
    async with httpx.AsyncClient() as client:
        resp = await client.get(url)
        return resp.text

Code-based tool

Register tools directly in Python:

from agent_gateway import Gateway

gw = Gateway(workspace="./workspace")

@gw.tool(agent="assistant")
def add_numbers(a: float, b: float) -> float:
    """Add two numbers together."""
    return a + b

Use the API

# Invoke an agent (single-turn)
curl -X POST http://localhost:8000/v1/agents/assistant/invoke \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{"message": "What is 2 + 3?"}'

# Chat with an agent (multi-turn)
curl -X POST http://localhost:8000/v1/agents/assistant/chat \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{"message": "Hello!"}'

Features

• Markdown-defined agents — Define agents, tools, and skills as markdown files with YAML frontmatter
• Multi-LLM support — Use any model supported by LiteLLM (OpenAI, Gemini, Anthropic, Ollama, etc.)
• Built-in authentication — API key and OAuth2/JWT auth out of the box
• Persistence — SQLite or PostgreSQL storage for conversations, executions, and audit logs
• Dashboard — Built-in web dashboard for monitoring agents, executions, and conversations
• Scheduling — Cron-based agent scheduling via APScheduler
• Notifications — Slack and webhook notification backends with per-agent rules
• Async execution — Queue-based async processing with Redis or RabbitMQ
• Telemetry — OpenTelemetry instrumentation with console or OTLP export
• Structured output — Pydantic model or JSON Schema output validation
• Agent memory — Automatic memory extraction and recall across conversations
• Streaming — Server-sent events (SSE) for real-time chat responses
• Input/output schemas — JSON Schema validation for agent inputs and outputs
• CLI — Project scaffolding, agent listing, and dev server via agents-gateway CLI
• Lifecycle hooks — before_invoke, after_invoke, on_error hooks for custom logic
• Sub-app mounting — Mount into an existing FastAPI app with gw.mount_to(app, path="/ai") — full feature parity

Sub-App Mounting

Mount the gateway into an existing FastAPI app with full feature parity — dashboard, auth, OAuth2, static assets, and all background subsystems work identically:

from fastapi import FastAPI
from agent_gateway import Gateway

app = FastAPI(title="My App")
gw = Gateway(workspace="./workspace")

gw.use_api_keys([{"name": "dev", "key": "secret", "scopes": ["*"]}])
gw.use_dashboard(auth_username="user", auth_password="pass",
                 admin_username="admin", admin_password="admin")

gw.mount_to(app, path="/ai")

# Your routes at /
# Gateway API at /ai/v1/...
# Dashboard at /ai/dashboard/

See the mounting guide for details.

Configuration

Configure your gateway with workspace/gateway.yaml:

server:
  port: 8000

model:
  default: "gemini/gemini-2.0-flash"
  temperature: 0.1

memory:
  enabled: true

Or configure programmatically:

from agent_gateway import Gateway

gw = Gateway(
    workspace="./workspace",
    title="My Agent Service",
)

# Fluent API for backends
gw.use_api_key_auth(api_key="your-key")
gw.use_sqlite("sqlite+aiosqlite:///data.db")
gw.use_slack_notifications(bot_token="xoxb-...", default_channel="#alerts")

Installation Extras

Install only what you need:

pip install agents-gateway[sqlite]       # SQLite persistence
pip install agents-gateway[postgres]     # PostgreSQL persistence
pip install agents-gateway[redis]        # Redis queue backend
pip install agents-gateway[rabbitmq]     # RabbitMQ queue backend
pip install agents-gateway[oauth2]       # OAuth2/JWT authentication
pip install agents-gateway[slack]        # Slack notifications
pip install agents-gateway[dashboard]    # Web dashboard
pip install agents-gateway[otlp]        # OTLP telemetry export
pip install agents-gateway[all]          # Everything

Documentation

Full documentation is available at vince-nyanga.github.io/agents-gateway.

License

MIT