Introduction to Datafuse

Welcome to Datafuse - the active integration studio and enterprise-ready tool-routing middleware designed specifically for AI agent systems.

Datafuse acts as a high-security gateway connecting your AI models, agents, and frameworks (like LangChain, LlamaIndex, CrewAI, or custom LLM stacks) to over 1,000+ API toolkits and managed user OAuth integrations with zero effort.

Rather than wasting weeks building authentication flows, coding API schema parsers, or handling user credential databases, Datafuse delivers a fully unified, decoupled, and sandboxed execution environment.

Why Teams Choose Datafuse

Datafuse is built from the ground up for high-velocity developer teams who require absolute credential safety, model agility, and dynamic integration control.

• Adding Custom Integrations: Instantly generate runtime-ready tool definitions by pasting OpenAPI/Swagger specifications or composing integration configurations in YAML.
• Decoupled Credentials Vault: Protect your users' sensitive credentials (OAuth tokens, API keys, basic authentication) from prompt injection risks. The AI agent only requests tool triggers; raw auth keys never touch agent prompts or models.
• 1,000+ Pre-Built Integrations: Tap into a massive directory of pre-compiled API connectors including Slack, GitHub, Gmail, Jira, Notion, Hubspot, and Salesforce.
• Bi-Directional Triggers: Establish robust event listeners that automatically stream external system activities (e.g. new email, git commit, webhook) back into your agent loops.
• Enterprise VPC & Compliance: Deploy Datafuse completely self-hosted inside your private virtual cloud (VPC) with cryptographically enforced zero-retention logs for rigorous SOC-2/HIPAA compliance.

High-Level Architecture

Datafuse sits securely as an active routing layer between your AI framework, the encrypted credentials vault, and the external world.

!IMPORTANT Because Datafuse decouples credential storage from tool definition, your agent framework receives standard, clean data responses. It is never exposed to raw client access tokens, protecting against dangerous prompt injection attacks designed to extract secrets.

Platform Core Runtimes

Datafuse supports three distinct integration runtime environments:

1. REST Connector: Static actions running over standard HTTP APIs with automated path parameter interpolation and custom authorization headers.
2. MCP (Model Context Protocol): Seamlessly bridge local and hosted MCP servers for real-time semantic search and context loading.
3. Database Connector: High-performance generated query interfaces supporting secure transactions against Postgres, SQLite, MySQL, and SQL Server.

Quick Navigation

Ready to connect your agent? Dive straight in:

• Installation Guide: Set up the Python and TypeScript SDKs.
• 5-Minute Quickstart: Use the v1 API to list integrations, resolve tools, and invoke a tool.
• 5-Minute Quickstart with curl: Run the same flow from the terminal with raw HTTP calls.
• Explore the Custom Integrations guide to compile bespoke tool schemas.
• Managed Authentication: Secure client logins with OAuth and custom vaults.