Context7 Docs
Give coding assistants current, version-specific library docs and code examples.
- Source repo
- upstash/context7
- Stars
- ★ 63k
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- 56/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer using a coding assistant such as Cursor wants documentation for the relevant Next.js version before generating proxy code.
- Not a fit if
- Teams that require every component to be self-hostable
- Teams that need guaranteed documentation accuracy or security
- Source review
- 56/100 · Major gaps
What does this agent do, and when should you use it?
Context7 is Upstash’s documentation retrieval platform for coding assistants, supplying current docs and examples from their sources. It offers two modes: the ctx7 CLI with a guiding skill, or an MCP server whose tools can be called by an MCP client. The CLI searches for libraries and retrieves their docs; the MCP tools are resolve-library-id and query-docs. Users can specify a Context7 library ID or version in a prompt to target the documentation they need. The repository contains the MCP server source and lists packages for a TypeScript SDK, Vercel AI SDK tools, and a pi.dev extension; the API backend, parsing engine, and crawling engine are private components.
In CLI mode, ctx7 library <name> <query> searches by library name and returns matching libraries and IDs, while ctx7 docs <libraryId> <query> retrieves docs by Context7-compatible library ID. In MCP mode, resolve-library-id finds an ID using libraryName and query, and query-docs retrieves documentation using libraryId and query; its optional language parameter ranks code examples for a language first. Retrieved documentation and examples are supplied to the coding assistant’s context. A prompt can include a library ID or version to target the lookup.
- A developer using a coding assistant such as Cursor wants documentation for the relevant Next.js version before generating proxy code.
- An engineer configuring Cloudflare Worker caching needs examples for caching JSON responses.
- A team building Supabase sign-up wants to retrieve the email and password registration API docs.
- A Node.js developer wants to search for a Context7 library ID and retrieve its docs from the terminal.
- A team with an MCP client wants its assistant to call library resolution and documentation retrieval tools natively.
How do you install or deploy this agent?
The CLI quickstart requires Node.js 22.13 or newer. Run setup to authenticate via OAuth and generate an API key, then choose CLI + Skills or MCP mode; flags can target a specific coding assistant. A free API key is recommended for higher rate limits. For manual MCP setup, configure the client with the Context7 server URL and pass a Bearer API key.
How do you use this agent?
After setup, search for a library and retrieve documentation with the CLI:
ctx7 library next.js "proxy configuration"
ctx7 docs /vercel/next.js "proxy configuration"Add use context7 to a prompt, or provide an exact library ID and version to target the docs. For MCP, configure the client with the server URL and Bearer key; the README does not specify a universal client configuration file format.
What are this agent's strengths and limitations?
- Both CLI and MCP paths retrieve current library docs and code examples.
- Library IDs and versions can be specified to reduce ambiguity in library matching.
- MCP queries accept a language parameter to rank matching code examples first.
- The project lists integration packages for the CLI, MCP server, TypeScript SDK, and Vercel AI SDK tools.
- The CLI requires Node.js 22.13 or newer.
- Retrieval depends on the Context7 service; manual MCP setup needs network access and an API key, while a free key is only documented as improving rate limits.
- The API backend, parsing engine, and crawling engine are private, so the full platform cannot be deployed from this repository alone.
- The project does not guarantee the accuracy, completeness, or security of community-contributed documentation.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Context7 Docs This agent | 56 · Major gaps | CLIFreemium | ★ 63k | today | TypeScript | — |
| Stock SDK | 71 · Some gaps | Library / SDKFree | ★ 2k | 17d ago | TypeScript | ChatGPT · Codex · Claude Code |
| Unbrowse Website API Client | 56 · Major gaps | Hosted serviceFreemium | ★ 802 | 3d ago | TypeScript | ChatGPT · Codex · Claude Code |
| Superset | 52 · Major gaps | Desktop appFreemium | ★ 15k | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The privacy document identifies data sent to the API, including formulated queries, library identifiers, API keys, client details, and transport type; it also discloses reranking and anonymous query storage. Tool descriptions instruct clients to remove sensitive material, supporting relatively strong data-flow and sensitive-data scores. Setup performs OAuth, generates an API key, and configures an agent, but the supplied files do not spell out granular permissions, confirmation details, credential-storage protections, or the external effects of each tool. A removal command provides limited rollback. The docs attribute indexed projects to their respective owners, and the README says backend, parsing, and crawling components are outside this repository, so attribution is traceable but incomplete.
The README distinguishes CLI and MCP modes and describes the tools, while the Ask API documentation lists status codes. However, the installation document is empty, and the privacy statement that full prompts are never transmitted lacks boundary detail about queries formulated from those prompts. Availability depends on APIs, hosted services, and external LLM providers, with no availability commitment in the supplied evidence. Ask status codes and retry guidance provide some failure handling.
The materials target coding agents retrieving library documentation and support CLI, MCP, multiple client setups, explicit library IDs, and version-specific requests. Privacy controls, source-type filters, and private deployment options document some capability boundaries. Invocation can be explicit or enabled through an agent rule. The supplied files do not cover all environments or failure configurations, so the criteria do not merit full marks.
The README organizes purpose, installation, tools, documentation, and packages; it includes a Node requirement, setup and removal instructions, multilingual entry points, and usage examples. Privacy boundaries and a security reporting path are documented, the MIT license text is complete, SECURITY.md lists supported versions and reporting steps, and workspace scripts plus canary publishing provide version-maintenance signals. The installation document is empty, and the supplied materials offer limited changelog history and no strong ongoing maintenance commitment, so scores remain below full.
The tools retrieve documentation relevant to a library and query, with examples showing reusable prompts and calls. Value depends on index quality and external services; the docs explicitly disclaim guarantees for community-contributed documentation accuracy, completeness, and security. Hosted reranking and query storage involve data processing, while the supplied materials do not quantify benefits, fees, or latency, limiting the cost-benefit score.
Tool names, parameters, setup commands, data fields, and API status codes are stated concretely, and some privacy claims include their stated uses. The README, privacy document, package metadata, and security policy corroborate the product structure and maintenance signals, but most service behavior is documented assertion; the repository explicitly omits backend and other components. The materials distinguish stated policy from documentation-quality limitations, but provide no independent support for all privacy and compliance claims.
- Queries are sent to the Context7 API and used for documentation reranking and anonymous benchmarking; avoid including secrets, credentials, or proprietary information.
- The repository omits the API backend, parsing engine, and crawling engine, so documentation accuracy, security, and service availability cannot be independently verified from these materials.
- The README requires Node.js 22.13 or newer, while the Docs7 quickstart says Node.js 20.19 or newer. These refer to different components; confirm the requirement for the tool you intend to use.