Context7

The Context7 Agent Plugin is a portable Agent Plugin (spec 1.0.0) designed to provide AI coding assistants with current, version-specific library documentation.

Instead of relying on potentially stale training data, the plugin uses the Context7 MCP server to fetch real documentation from source repositories at query time.

Included Components:

• MCP Server: A remote server over Streamable HTTP, authorized via OAuth (configured in mcp.json).
• Skill: A documentation lookup trigger (located in skills/context7-mcp/) that activates when you ask the agent about a specific library.

Context7 implements several security measures to protect user data and ensure privacy during documentation retrieval:

• Prompt Privacy: Your original prompts remain with your AI assistant. Context7 only receives search queries formulated by the MCP client, which is instructed to strip sensitive data before transmission.
• Infrastructure: Documentation is indexed within SOC 2 compliant infrastructure operated by Upstash.
• API Security: API keys are encrypted, rate-limited, and can be rotated via the Upstash dashboard.
• Enterprise Features: Enterprise customers have access to SSO (SAML, OAuth, OIDC) and dedicated audit trails.

For full legal and compliance details, refer to the Context7 Addendum and Upstash Terms and Privacy Policy.

Context7 maintains transparency through its open-source model and leverages Upstash's compliance framework:

Open Source

The Context7 MCP server is open source. The code is publicly available on GitHub for community auditing and contribution.

• Repository: github.com/upstash/context7

Compliance Certifications

Context7 benefits from Upstash's existing compliance certifications:

• SOC 2 Type II
• GDPR compliant
• ISO 27001 (currently in progress)

Context7 can be integrated into your workflows via several different interfaces:

• Vercel AI SDK: Use generateText, streamText, or the pre-built Context7Agent for automatic documentation lookup.
• TypeScript SDK: Call Context7 directly from any TypeScript or Node.js application.
• MCP Server: Connect the Context7 Model Context Protocol (MCP) server to any compatible agent or editor.

The Context7 Plugin for Cursor addresses the issue of outdated training data and hallucinated APIs in AI coding assistants. It enables Cursor to fetch real-time, up-to-date documentation directly from source repositories instead of relying on stale model knowledge.

The plugin provides several key components:

• MCP Server: Connects Cursor to the Context7 documentation service.
• Rules: An always-on use-context7 rule that prompts the agent to fetch documentation when it is uncertain about library APIs.
• Skills: A context7-docs-lookup skill containing instructions for resolving libraries and fetching documentation.
• Agents: A dedicated docs-researcher agent designed for focused documentation lookups.

Context7 MCP is a Model Context Protocol (MCP) server that provides up-to-date, version-specific documentation and code examples directly to your LLM. It prevents hallucinations and outdated code generation by fetching real-time data from the source.

To use it, include use context7 in your prompt. This tells the LLM to invoke the MCP tool to fetch the necessary context for your request.

Example Prompts:

• Create a Next.js middleware that checks for a valid JWT in cookies and redirects unauthenticated users to /login. use context7
• Configure a Cloudflare Worker script to cache JSON API responses for five minutes. use context7

Create a Next.js middleware that checks for a valid JWT in cookies
and redirects unauthenticated users to `/login`. use context7

The documentation lookup follows a two-step workflow orchestrated by the AI model:

1. resolveLibraryId: Searches the Context7 database to find the correct library ID for a query (e.g., converting "react" to /reactjs/react.dev).
2. queryDocs: Uses the resolved library ID and the user's query to fetch specific, relevant documentation content.

This allows the model to fetch current documentation on-demand rather than relying on potentially outdated training data.

Context7 uses a layered malicious content detection system to ensure that documentation retrieved from public or private sources is safe for human developers and AI coding agents to consume. The system includes:

• Robust Detection: Uses a specialized classifier tailored for Context7 to identify prompt injection attempts and malware-related patterns.
• Targeted Validation: Suspicious content undergoes additional validation checks.
• Continuous Monitoring: Flagged content is tracked and reviewed continuously.
• Regular Updates: Detection logic is updated frequently to defend against evolving attack methods and new injection techniques.

By default, Context7 On-Premise runs as a single container using SQLite and LanceDB. To scale horizontally with multiple replicas behind a load balancer, you must move state to an external PostgreSQL database with the pgvector extension.

Once DATABASE_URL is configured, every replica becomes stateless and interchangeable. Replicas coordinate via Postgres to pull indexing jobs from a shared queue and use Postgres locks for singleton tasks (like backups or GitOps sync). Adding replicas increases both query capacity and indexing throughput simultaneously.

# Architecture Overview
Clients -> Load Balancer -> Multiple Replicas -> PostgreSQL + pgvector