Superglue

Superglue is an agentic integration platform where AI agents write, monitor, and repair production-grade enterprise integrations from natural language prompts. Rather than maintaining a library of pre-built connectors — the dominant model in the iPaaS space — Superglue takes a documentation-driven approach: it reads API documentation, understands custom authentication patterns, handles rate limiting, and navigates the misleading HTTP 200 status codes that frequently plague legacy system integrations.

At its architectural core, Superglue operates on two primitives: Systems and Tools. Systems are reusable, configurable containers representing any data source — REST APIs, GraphQL endpoints, SOAP services, SFTP servers, PostgreSQL databases, and Redis instances — each configured with documentation, URLs, and authentication tailored to the use case. Tools are AI-built execution pipelines that run on these systems, processing inputs through sequential steps and returning structured outputs within Superglue's managed runtime. The platform handles the entire data pipeline: making API calls, resolving credentials, managing retries and rate limiting, and logging every execution.

Superglue exposes tools through multiple access points: a REST API, TypeScript and Python SDKs, Model Context Protocol (MCP) endpoints for AI agents, a CLI with built-in agent skills covering 35+ AI coding assistants, and a web dashboard. For automated workflows, tools can be triggered via cron-based schedules and incoming webhooks, with every execution logged for debugging and audit purposes.

Self-hosted deployment requires Docker, a PostgreSQL database, and an LLM API key from OpenAI, Anthropic, or Gemini. File storage via S3 or MinIO is needed for documentation-scraping features such as URL fetching, OpenAPI spec ingestion, and file uploads — without it, systems lack the documentation context the AI agent relies on. A cloud-hosted option is available at api.superglue.cloud, offering the same API and MCP interfaces without infrastructure management.

The CLI experience deserves particular attention. The superglue CLI (sg) ships with a skill — a structured reference file that teaches AI coding agents how to use every sg command, handle authentication, build tools, and debug failures. This skill works across 35+ AI coding agents, allowing developers to integrate Superglue into their existing AI-assisted development workflows without additional configuration. Common workflows include building tools from natural language descriptions, debugging failed executions with step-level result inspection via --include-step-results, and deploying tools through SDK, REST API, or webhook patterns.

The MCP integration is a key architectural feature. By appending /mcp to the API endpoint, Superglue exposes all configured tools as MCP servers that AI agents can discover and invoke. Export configurations are available for Langdock, Claude Code, Cursor, and Codex, positioning Superglue as a bridge between enterprise systems and the growing ecosystem of MCP-compatible AI tools.

Superglue occupies a distinct position in the integration landscape. It is neither a traditional iPaaS with a fixed connector catalog nor a pure AI agent platform — it is an execution environment where AI agents both build and run integrations. For teams working within the AI Agent Development ecosystem or using agent infrastructure like Openclaw, Superglue's MCP-native design makes it a natural extension for connecting AI agents to real-world systems. Organizations evaluating alternative approaches such as Genspark.ai may find Superglue's documentation-driven model more suitable for environments with many legacy or custom APIs that lack pre-built connectors.