grip serve - Grip AI

Overview

The grip serve command starts a lightweight REST API server without the full gateway features. It provides HTTP endpoints for agent interaction but skips channels (Telegram, Discord, Slack), cron scheduler, and heartbeat services. Use this when:

You only need the HTTP API
You don’t want messaging platform integrations
You want a minimal deployment
You’re building custom integrations

Usage

grip serve [OPTIONS]

What It Starts

REST API Server

HTTP API with FastAPI + Uvicorn:

Chat endpoints - /api/chat for agent conversations
Session management - /api/sessions for session control
Tool execution - /api/tools for direct tool calls
Health checks - /api/health for monitoring
MCP servers - /api/mcp for server management

Default: http://0.0.0.0:8080

What It Excludes

Compared to grip gateway, this does NOT start:

❌ Telegram, Discord, or Slack channels
❌ Cron scheduler for periodic tasks
❌ Heartbeat service for autonomous check-ins
❌ Message bus for channel routing

Options

—host / -H

Override the bind address from config:

grip serve --host 127.0.0.1
grip serve -H 0.0.0.0

Default: 0.0.0.0 (all interfaces) Common values:

0.0.0.0 - Listen on all network interfaces
127.0.0.1 - Localhost only (most secure)
192.168.1.100 - Specific interface

—port / -p

Override the bind port from config:

grip serve --port 3000
grip serve -p 8080

Default: 8080

Starting the Server

Basic Start

$ grip serve

grip API server starting on 0.0.0.0:8080

Output indicates:

Server is running
Bind address and port
Ready to accept requests

Custom Port

$ grip serve --port 3000

grip API server starting on 0.0.0.0:3000

Localhost Only

$ grip serve --host 127.0.0.1

grip API server starting on 127.0.0.1:8080

Custom Host and Port

$ grip serve --host 127.0.0.1 --port 3000

grip API server starting on 127.0.0.1:3000

API Endpoints

Once running, access these endpoints:

Health Check

curl http://localhost:8080/health

Response:

{
  "status": "healthy",
  "version": "1.0.0",
  "uptime_seconds": 120
}

Chat API

curl -X POST http://localhost:8080/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "What time is it?",
    "session_id": "api:test-session"
  }'

Response:

{
  "response": "The current time is 2:47 PM.",
  "session_id": "api:test-session",
  "iterations": 1,
  "tokens_used": 89
}

Session Management

# List sessions
curl http://localhost:8080/api/sessions

# Get session details
curl http://localhost:8080/api/sessions/api:test-session

# Clear session
curl -X DELETE http://localhost:8080/api/sessions/api:test-session

Tool Execution

curl -X POST http://localhost:8080/api/tools/bash \
  -H "Content-Type: application/json" \
  -d '{
    "command": "ls -la"
  }'

MCP Server Management

# List MCP servers
curl http://localhost:8080/api/mcp/servers

# Get server tools
curl http://localhost:8080/api/mcp/servers/todoist/tools

Configuration

The server reads from ~/.grip/config.json:

{
  "gateway": {
    "host": "0.0.0.0",
    "port": 8080,
    "api": {
      "max_request_body_bytes": 1048576,
      "rate_limit_per_minute": 60,
      "rate_limit_per_minute_per_ip": 10
    }
  }
}

Key settings:

host - Bind address (overridden by --host)
port - Bind port (overridden by --port)
max_request_body_bytes - Maximum request size (default: 1MB)
rate_limit_per_minute - Global rate limit
rate_limit_per_minute_per_ip - Per-IP rate limit

Stopping the Server

Press Ctrl+C to stop:

^C
Shutting down...

The server shuts down gracefully:

Stops accepting new requests
Completes in-progress requests
Closes connections
Exits cleanly

Always use Ctrl+C or kill <pid> (not kill -9) for graceful shutdown.

Dependencies

The API server requires optional dependencies:

# Install with pip
pip install grip[api]

# Or with uv
uv pip install grip[api]

If not installed:

$ grip serve

API dependencies not installed.
Install with: pip install grip[api]  or  uv pip install grip[api]

Included packages:

FastAPI - Web framework
Uvicorn - ASGI server
Pydantic - Request validation
Python-multipart - File uploads

Use Cases

API-Only Deployment

Run Grip AI as a backend service:

grip serve --host 0.0.0.0 --port 8080

Access from your application:

import requests

response = requests.post(
    "http://your-server:8080/api/chat",
    json={
        "message": "Analyze this data",
        "session_id": "user_123"
    }
)
result = response.json()

Development Server

Quick local testing without channels:

grip serve --host 127.0.0.1 --port 3000

Microservice Architecture

Embed Grip AI in a larger system:

# Grip AI service
grip serve --port 8080

# Your frontend (port 3000)
# Your database (port 5432)
# Other services...

Docker Container

FROM python:3.11-slim
WORKDIR /app
COPY . .
RUN pip install grip[api]
EXPOSE 8080
CMD ["grip", "serve", "--host", "0.0.0.0"]

docker build -t grip-api .
docker run -p 8080:8080 grip-api

Production Deployment

Using systemd

Create /etc/systemd/system/grip-serve.service:

[Unit]
Description=Grip AI API Server
After=network.target

[Service]
Type=simple
User=grip
WorkingDirectory=/home/grip
ExecStart=/usr/local/bin/grip serve
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

Commands:

sudo systemctl enable grip-serve
sudo systemctl start grip-serve
sudo systemctl status grip-serve

Behind Nginx

server {
    listen 80;
    server_name api.example.com;

    location / {
        proxy_pass http://127.0.0.1:8080;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
    }
}

grip serve --host 127.0.0.1 --port 8080

Using PM2

pm2 start "grip serve" --name grip-api
pm2 save
pm2 startup

Docker Compose

version: '3.8'
services:
  grip-api:
    image: grip-api
    ports:
      - "8080:8080"
    volumes:
      - ~/.grip/config.json:/root/.grip/config.json
      - ~/grip/workspace:/root/grip/workspace
    restart: always

docker-compose up -d

Monitoring

Health Checks

# Simple check
curl http://localhost:8080/health

# With timeout
curl --max-time 5 http://localhost:8080/health

# In a loop
while true; do
  curl -s http://localhost:8080/health | jq .
  sleep 60
done

Logs

Uvicorn logs are printed to stdout:

grip serve 2>&1 | tee server.log

Metrics

Access logs show request patterns:

INFO:     127.0.0.1:52342 - "POST /api/chat HTTP/1.1" 200 OK
INFO:     127.0.0.1:52343 - "GET /health HTTP/1.1" 200 OK

Troubleshooting

Port Already in Use

$ grip serve

Error: [Errno 48] Address already in use

Solution:

# Find process using port
lsof -i :8080

# Kill process or use different port
grip serve --port 3000

API Dependencies Missing

$ grip serve

API dependencies not installed.

Solution:

pip install grip[api]
# or
uv pip install grip[api]

Connection Refused

$ curl http://localhost:8080/health
curl: (7) Failed to connect to localhost port 8080

Check:

Server is running
Correct host and port
Firewall settings
Server logs for errors

High Memory Usage

# Check memory
ps aux | grep grip

# Limit workers
grip serve --workers 1

# Clear old sessions
find ~/grip/workspace/sessions/ -mtime +7 -delete

Comparison: serve vs gateway

Feature	`grip serve`	`grip gateway`
REST API	✓	✓
Telegram	✗	✓
Discord	✗	✓
Slack	✗	✓
Cron scheduler	✗	✓
Heartbeat service	✗	✓
Message bus	✗	✓
Startup time	Fast	Slower
Memory usage	Low	Higher
Use case	API-only	Full platform

When to use grip serve:

Building custom integrations
Embedding in applications
API-only deployments
Minimal resource usage
Development/testing

When to use grip gateway:

Need messaging platforms
Want scheduled tasks
Require autonomous agents
Full feature set
Production deployments

grip gateway - Full platform with channels and services
grip agent - Interactive chat (no server required)
grip config - Configure server settings
grip status - Check system status

REST API

CLI Commands

Built-in Tools

​Overview

​Usage

​What It Starts

​REST API Server

​What It Excludes

​Options

​—host / -H

​—port / -p

​Starting the Server

​Basic Start

​Custom Port

​Localhost Only

​Custom Host and Port

​API Endpoints

​Health Check

​Chat API

​Session Management

​Tool Execution

​MCP Server Management

​Configuration

​Stopping the Server

​Dependencies

​Use Cases

​API-Only Deployment

​Development Server

​Microservice Architecture

​Docker Container

​Production Deployment

​Using systemd

​Behind Nginx

​Using PM2

​Docker Compose

​Monitoring

​Health Checks

​Logs

​Metrics

​Troubleshooting

​Port Already in Use

​API Dependencies Missing

​Connection Refused

​High Memory Usage

​Comparison: serve vs gateway

​Related Commands

Build docs developers (and LLMs) love

Overview

Usage

What It Starts

REST API Server

What It Excludes

Options

—host / -H

—port / -p

Starting the Server

Basic Start

Custom Port

Localhost Only

Custom Host and Port

API Endpoints

Health Check

Chat API

Session Management

Tool Execution

MCP Server Management

Configuration

Stopping the Server

Dependencies

Use Cases

API-Only Deployment

Development Server

Microservice Architecture

Docker Container

Production Deployment

Using systemd

Behind Nginx

Using PM2

Docker Compose

Monitoring

Health Checks

Logs

Metrics

Troubleshooting

Port Already in Use

API Dependencies Missing

Connection Refused

High Memory Usage

Comparison: serve vs gateway

Related Commands