Skip to content
Scalekit Docs
Talk to an Engineer Dashboard

Use Scalekit docs in your AI coding agent

Use Context7 to give your AI coding agent accurate, up-to-date Scalekit documentation so it can help you integrate faster and with fewer errors.

AI coding agents like Claude Code and Cursor work from training data that can be months out of date. When you ask them to help integrate Scalekit, they may reference old APIs, deprecated patterns, or incorrect parameter names — leading to bugs that are hard to trace.

Context7 is an MCP (Model Context Protocol) server that pulls live, version-accurate documentation directly into your coding agent’s context. It is opt-in: docs are only injected when you add use context7 to a prompt, or when you configure auto-invoke rules so the agent calls it automatically. Without one of these triggers, Context7 is not invoked and your agent falls back to its training data.

Scalekit’s full developer documentation is indexed on Context7 at context7.com/scalekit-inc/developer-docs, covering hundreds of pages and thousands of code snippets across SSO, SCIM, MCP auth, agent auth, and connected accounts.

Get accurate answers about Scalekit

Section titled “Get accurate answers about Scalekit”

When a prompt triggers Context7 — either via the use context7 keyword or an auto-invoke rule — it retrieves the relevant documentation from the indexed Scalekit docs and injects it into the agent’s context window. Your agent then answers using accurate, current content rather than training data.

Context7 provides two MCP tools:

  • resolve-library-id — finds the Scalekit docs library by name
  • query-docs — retrieves specific documentation sections by library ID

  1. Set up Context7

    Section titled “Set up Context7”

    Context7 is configured as an MCP server in your coding agent. You can also add it directly from context7.com. Choose your tool:

    Run one of the following commands in your terminal:

    Local (stdio):

    Terminal window
    claude mcp add --scope user context7 -- npx -y @upstash/context7-mcp

    Remote (HTTP):

    Terminal window
    claude mcp add --scope user --transport http context7 https://mcp.context7.com/mcp

    To verify the server was added:

    Terminal window
    claude mcp list

  2. Query Scalekit docs

    Section titled “Query Scalekit docs”

    Once Context7 is running, add use context7 to any prompt where you want current Scalekit documentation injected automatically.

    General Scalekit queries:

    How do I set up SSO with Scalekit? use context7
    Show me how to provision users with SCIM using Scalekit. use context7

    Target Scalekit docs directly using the library path:

    use library /scalekit-inc/developer-docs for how to add MCP auth to my server

    Combine with version or feature specificity:

    How do I configure connected accounts for GitHub OAuth with Scalekit? use context7

  3. Auto-invoke Context7 (optional)

    Section titled “Auto-invoke Context7 (optional)”

    Configure your coding agent to always use Context7 for library and API questions — no need to add “use context7” manually each time.

    Add the following rule to your project’s CLAUDE.md file:

    Always use Context7 MCP when I need library or API documentation, code generation, or setup and configuration steps.

    This applies project-wide. For a global rule, add it to ~/.claude/CLAUDE.md.

  4. Increase rate limits with an API key

    Section titled “Increase rate limits with an API key”

    The free tier of Context7 has rate limits. For heavier usage or team environments, get a free API key from context7.com/dashboard and add it to your configuration.

    Local:

    Terminal window
    claude mcp add --scope user context7 -- npx -y @upstash/context7-mcp --api-key YOUR_API_KEY

    Remote:

    Terminal window
    claude mcp add --scope user --header "CONTEXT7_API_KEY: YOUR_API_KEY" --transport http context7 https://mcp.context7.com/mcp

    API keys start with ctx7sk. If authentication fails with a 401 error, verify the key format matches your transport method (HTTP header vs. CLI argument).