Prodigent

Help

Connect with MCP

Prodigent exposes a remote Model Context Protocol server so assistants like Claude and Cursor can search and capture your organisation’s institutional knowledge — without leaving the tools where you already work.

What you get

Once connected, the assistant authenticates as you and operates inside your current Prodigent workspace. It can search across entities, capture knowledge, ask questions, manage incidents, and work with initiatives and solution designs.

How authentication works

Prodigent’s MCP endpoint uses OAuth. When you connect from Claude, Cursor, or another compatible client, you’ll be redirected to sign in and approve access. You don’t need to copy API keys for the standard flow — leave OAuth Client ID / Secret blank unless your organisation’s MCP client requires pre-registered credentials.

Connect Claude (claude.ai)

Custom remote MCP connectors work on Free, Pro, Max, Team, and Enterprise plans (Free is limited to one custom connector). Claude reaches the server from Anthropic’s cloud, so the URL above must be publicly reachable.

Pro and Max

  1. 1

    Open Connectors

    Go to Customize → Connectors (or open claude.ai/settings/connectors ).

  2. 2

    Add a custom connector

    Click +, then Add custom connector. Name it something like “Prodigent” and paste the server URL above.

  3. 3

    Connect and approve

    Finish with Add, then Connect. Sign in to Prodigent and approve the request when prompted.

  4. 4

    Enable it in a chat

    In a conversation (including inside a Claude Project), open + → Connectors and toggle Prodigent on for that chat.

Team and Enterprise

An Owner adds the connector under Organization settings → Connectors (Add → Custom → Web, using the same URL. Each member then connects individually from Customize → Connectors so Claude only uses that person’s Prodigent access.

Steps follow Anthropic’s custom connectors guide . UI labels can shift slightly; the server URL stays the same.

Add Prodigent to a Claude Code project

For team coding workflows, register Prodigent at project scope. That writes a .mcp.json at the repo root so everyone who clones the project gets the same connector.

  1. 1

    Add the server from the project root

    claude mcp add --transport http --scope project prodigent https://app.prodigent.ai/mcp/prodigent
  2. 2

    Commit .mcp.json

    Review and commit the generated file so the team shares the same MCP config. Each person still authenticates with their own Prodigent account.

  3. 3

    Approve and connect

    On first use, Claude Code asks you to approve the project MCP server. Run /mcp if you need to finish OAuth or check connection status.

Prefer not to commit config? Omit --scope project for a private local install, or use --scope user to enable Prodigent across all of your Claude Code projects. 

{
  "mcpServers": {
    "prodigent": {
      "type": "http",
      "url": "https://app.prodigent.ai/mcp/prodigent"
    }
  }
}

Example .mcp.json shape. Exact keys may follow your Claude Code version — prefer claude mcp add when unsure.

Connect Cursor

In Cursor, add a remote MCP server that points at the same URL. Cursor will run the OAuth flow against Prodigent (including dynamic client registration). After you approve access, Prodigent tools appear alongside your other MCP servers.

Available tools

These tools are available on the Prodigent MCP server today. Prefer search tools to resolve slugs before fetching history or designs.

Tool Capability
search Broad search across knowledge, components, terminology, docs, runbooks, incidents, and initiatives.
search_knowledge Search captured knowledge notes in your workspace by keyword.
capture_knowledge Record institutional knowledge — what someone should know — with an optional component attachment and source URL.
search_components Find components (services, applications, databases, and other types) by name, description, or terminology.
search_terminology Search aliases and other confirmed terminology and see what each term refers to.
search_docs Search knowledge documents by title, description, or content.
search_runbooks Search runbooks by title, description, owner, or terminology.
search_incidents Search incidents by title, summary, impact, detection, resolution, or terminology.
upsert_incident Create a new incident or update an existing one (by id or slug) — title, summary, impact, detection, resolution, status.
ask Ask a natural-language question against workspace knowledge. Returns an AI answer when available, with referenced notes and components.
search_initiatives Find initiatives by title, slug, or terminology. Use this to resolve a slug before fetch/design/question tools.
fetch_initiative Fetch initiative details including PRD excerpt, requirements, linked components, and design overview.
ask_initiative_question Record an open question on an initiative’s solution design.
get_solution_design Return an initiative’s solution design as markdown, plus open questions, standing decisions, and linked components.

Suggested usage

  • Use search for broad discovery, then typed search_* tools when you know the entity kind.
  • Run search_knowledge before capturing something that might already exist.
  • Use ask for natural-language questions over workspace knowledge.
  • Resolve initiatives with search_initiatives, then fetch_initiative, get_solution_design, or ask_initiative_question.

Need a Prodigent workspace first?

Try Prodigent free All help topics