Agent Integration

Loop integrates with AI agents through multiple surfaces. Choose the one that fits your workflow, or follow a per-agent guide below.

Decision Tree

Use this diagram to pick the right integration surface for your use case.

graph TD
    A["Integrating an AI agent with Loop?"] --> B{"Want zero-code setup<br/>in your IDE?"}
    B -->|Yes| C["MCP Server<br/>/docs/mcp"]
    B -->|No| D{"Need contextual<br/>instructions for your agent?"}
    D -->|Yes| E["Agent Skill<br/>/docs/skill"]
    D -->|No| F{"Building a custom<br/>application?"}
    F -->|Yes| G["TypeScript SDK<br/>/docs/sdk"]
    F -->|No| H{"Automating from<br/>scripts or CI?"}
    H -->|Yes| I["CLI<br/>/docs/cli"]
    H -->|No| J["REST API<br/>/docs/api"]

Per-Agent Guides

Step-by-step setup instructions for specific AI agents and IDEs.

Integration Surfaces

Loop provides five integration surfaces. Most setups combine two or more.

MCP Server — Zero-code IDE integration via the Model Context Protocol. Install in one command, and your agent gets 9 tools for claiming tasks, creating issues, ingesting signals, and more.

TypeScript SDK — Programmatic access to Loop for custom applications, agent frameworks, and scripts. Type-safe client with resource-based API.

Agent Skill — Contextual instructions packaged as a content-only npm module. Teaches AI agents how to use Loop effectively by providing a SKILL.md (or equivalent) that the agent reads for context.

CLI — Command-line interface for managing issues, signals, triage, templates, and dispatch from scripts or CI pipelines.

REST API — Direct HTTP access for any language or platform. Every other surface (MCP, SDK, CLI) communicates through this API.