Connect your AI

Your training data, wherever you work.

repClimb speaks MCP, an open standard that lets AI clients securely read your workouts, understand your progress, and help shape your next plan.

Streamable HTTP endpoint

https://www.repclimb.app/api/mcp/mcp

OAuth protectedRead tools includedWrites need mcp:write

The short version

Three steps to get moving

01

Add the endpoint

Choose a client below and paste the endpoint into its remote MCP server settings.

02

Sign in

repClimb opens Clerk OAuth in your browser. Approve access for the account whose data you want to use.

03

Ask a question

Try “What did I bench last time?” or “Show me my squat progress.” Your agent can now use your real history.

Scope-aware access: mcp:read lists and enables the four read tools; mcp:write independently lists and enables the four write tools. A combined grant is required to use both surfaces. Tokens without an explicit repClimb scope do not receive read access.
First connection? Sign in or create your repClimb account in the OAuth window opened by your MCP client, then approve the requested mcp:read and/or mcp:write scopes. That OAuth approval is the external data-sharing approval; no separate repClimb cookie or consent page is required.

Choose your client

Cx

OpenAI

Codex

Codex CLI, the Codex IDE extension, and the ChatGPT desktop app share the same MCP configuration. The CLI is the fastest setup.

Terminal
codex mcp add repclimb --url https://www.repclimb.app/api/mcp/mcp
codex mcp login repclimb
codex mcp list

In the Codex TUI, type /mcp to inspect connected servers. You can also open Settings → MCP servers in the ChatGPT desktop app or IDE extension, choose Streamable HTTP, and authenticate.

Verify: run codex mcp list, then ask Codex to list your latest repClimb workout.
CC

Anthropic

Claude Code

Claude Code supports remote HTTP MCP servers and OAuth. Add repClimb to your current project or use the user scope to make it available across projects.

Terminal
claude mcp add --transport http repclimb https://www.repclimb.app/api/mcp/mcp
claude mcp login repclimb
claude mcp list

Start Claude Code, then use /mcp to see the server and clear or repeat its OAuth login. To share the setup across every project, add --scope user when adding it.

Verify: run claude mcp list and look for a connected repclimb server.
CW

Anthropic

Claude web

Claude web connects to remote MCP servers through custom connectors. This works from the browser and keeps the same repClimb OAuth flow as Claude Code.

  1. 01 Open Settings → Connectors in Claude.
  2. 02 Choose Add custom connector and paste the repClimb endpoint.
  3. 03 Finish the repClimb OAuth sign-in and approve access.
  4. 04 Enable the connector in a chat, then ask about your training history.
Verify: ask Claude, “What did I bench press last time?” and confirm it uses the repClimb connector.

Claude web connects from Anthropic's cloud, so use the public HTTPS endpoint above. Team and Enterprise workspaces may require an owner or admin to approve the connector first. Claude connector help ↗

G

OpenAI

ChatGPT desktop

The desktop app has a visual MCP server setup for a custom remote server. Add repClimb, authenticate with OAuth, and use the tools from a ChatGPT conversation.

  1. 01 Open Settings → MCP servers.
  2. 02 Select Add server and choose Streamable HTTP.
  3. 03 Paste the endpoint and save. Restart the app if prompted.
  4. 04 Select Authenticate, finish sign-in, then type /mcp.
Verify: ask ChatGPT, “Show my last three workouts,” and approve the repClimb tool call when prompted.
GW

OpenAI

ChatGPT web

ChatGPT web can use repClimb through a custom MCP app. The current OpenAI flow is intended for Business and Enterprise/Edu workspaces with Developer Mode enabled.

  1. 01 Ask a workspace owner or admin to enable Developer Mode for ChatGPT apps.
  2. 02 Create a custom app and enter the repClimb MCP endpoint.
  3. 03 Scan the available tools, choose the OAuth option, and complete repClimb sign-in.
  4. 04 Publish the app, select it from the chat tools menu, and ask about your workouts.
Web note: ChatGPT's custom MCP app flow is currently a workspace/developer-mode setup, not a regular personal-chat “paste URL” setting. Long-lived OAuth access may require refresh-token support and the offline_access scope. ChatGPT MCP app help ↗
Cu

Cursor

Cursor

Cursor reads project-level MCP configuration from .cursor/mcp.json or global configuration from ~/.cursor/mcp.json. OAuth is handled from Cursor’s MCP settings.

.cursor/mcp.json
{
  "mcpServers": {
    "repclimb": {
      "url": "https://www.repclimb.app/api/mcp/mcp"
    }
  }
}

Save the file, then open Settings → Features → MCP. When repClimb shows an authentication action, use it to complete the browser sign-in.

Verify: the repClimb server should show as connected and its tools should appear under Available Tools.
Ge

Google

Gemini CLI

Gemini CLI can add an HTTP MCP server from the terminal and exposes OAuth through its interactive MCP commands.

Terminal
gemini mcp add repclimb https://www.repclimb.app/api/mcp/mcp --transport http
gemini mcp list

Start Gemini CLI and run /mcp auth repclimb to complete OAuth. Use /mcp list or gemini mcp list to inspect the connection.

Verify: ask Gemini to use the repClimb tools, for example: “Show my last three workouts.”
M

Testing

MCP Inspector

Use the Inspector when you want to see the raw handshake, available tools, and OAuth behavior before connecting an AI client.

Terminal
npx @modelcontextprotocol/inspector https://www.repclimb.app/api/mcp/mcp
Verify: initialize the connection, authenticate when prompted, and confirm that tools such as query_workouts and get_exercise_stats are listed.

What repClimb exposes

Useful context, with guardrails.

Read your history

Four read tools cover workouts and optional sets, the exercise catalog, training configuration, and exercise stats.

Plan with context

Four write tools create/edit drafts, finalize a reviewed plan, and start a planned workout.

Your identity stays yours

Every request is scoped to the account authenticated through OAuth.

Scopes are explicit

mcp:read and mcp:write are independent; review the external client before approving writes.

Codex MCP docs ↗Claude Code MCP docs ↗Cursor MCP docs ↗Gemini CLI MCP docs ↗