mcpField
T
Tally Workout

Tally Workout

Remote
tallyworkout.appHealth0Updated 2w ago

About Tally Workout

Tally Workout is a remote MCP server for the Tally iPhone strength log. It lets an AI agent review training records and plan workouts in the app, with user-requested changes to completed workouts that can be restored for seven days.

Key features

  1. 01Read workout history and training context
  2. 02Review exercise progress and weekly summaries
  3. 03Create and edit planned workout days
  4. 04Log, update, delete, or restore workouts
  5. 05Search exercises and saved templates

Use cases

  1. 01Plan upcoming sessions from recent training
  2. 02Review progress and training volume
  3. 03Record a completed workout through an agent

MCP support

Exposes 21 read and write tools through a remote Streamable HTTP MCP endpoint. Authentication uses OAuth 2.1 or an API key, with access levels governing workout changes.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.tallyworkout.app/api/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://mcp.tallyworkout.app/api/mcpAuth required

Tools

Details

From the project README

Tally Workout MCP server

Docs for the remote Model Context Protocol server of Tally Workout, a free iPhone strength log. Your AI agent reads your training history and plans workouts in the Tally app. You log every set at the gym. Next week, the agent plans from what you actually lifted. When you ask, the agent can also log, change or delete a finished workout, and you can undo each change for 7 days.

It works with Claude, Claude Code, Codex, ChatGPT, Muse, Hark, Perplexity, Dots and any MCP client that supports OAuth 2.1 or a bearer token.

This repo holds the public docs. The server is first-party and closed source. It runs as part of the Tally backend.

Endpoint

https://mcp.tallyworkout.app/api/mcp
  • Transport: Streamable HTTP. No SSE-only mode.
  • Auth, option 1: OAuth 2.1, authorization code + PKCE (S256), public client, dynamic client registration (RFC 7591). Scope: workouts.
  • Auth, option 2: an API key, for clients without OAuth. Create it in the Tally app under Settings → API keys. Send it as Authorization: Bearer tally_ct_…. A key is Read only, Read and plan, or Full access. Only a Full access key or an OAuth connection can change finished workouts. Keys do not expire. Revoke a key in the app.
  • Discovery: /.well-known/mcp-server, agent brief (llms.txt).

Setup by agent

You need the Tally iPhone app and a Tally account. Each link has the full steps.

AgentAuthSteps
Claude (web, desktop, phone)OAuthAdd Tally Workout from Claude's directory. Steps
Claude CodeOAuth or API keyclaude mcp add --scope user --transport http tally https://mcp.tallyworkout.app/api/mcp, then /mcp. Steps
CodexAPI keyThree lines in ~/.codex/config.toml. Steps
ChatGPTOAuthDeveloper mode. Pro reads only. Business, Enterprise and Edu can plan. Steps
MuseOAuthClient ID tally-muse. Steps
HarkAPI keySteps
PerplexityAPI keySteps
DotsAPI keyNot tested yet. Steps
Any MCP clientOAuth or API keySteps

Tools

21 tools: 10 read, 11 write.

ToolTypeWhat it does
get_training_contextreadReturns planned days from today, the weight unit and, with historyWeeks, recent workouts.
get_workout_historyreadReturns logged workouts, newest first, set by set. Each workout has a stable id.
get_workoutreadReturns one logged workout by its id, set by set.
get_exercise_progressreadReturns progress for up to 10 exercises over the last 365 days, with top sets and estimated 1RM.
get_planned_workoutreadReturns one planned day in full, with its version.
search_exercisesreadMatches exercise names to the user's recent exercises and the Tally library.
get_exercisereadReturns one exercise by its id: muscles, equipment and instructions, or the user's custom exercise.
get_training_summaryreadReturns weekly totals for up to 12 weeks: workouts, sets, volume, time, volume by muscle, and missed or skipped plans.
get_saved_workoutsreadReturns the user's saved and favorite workout templates.
get_workout_versionsreadLists the restorable versions of one finished workout, also a deleted one.
plan_workout_daywriteCreates or replaces one planned day.
plan_workout_rangewriteCreates or replaces up to 31 planned days in one call.
update_planned_workoutwriteChanges part of one planned day: add, remove or change exercises.
move_planned_workoutwriteMoves a planned day to an empty date.
skip_planned_workoutwriteMarks a planned day as skipped, or restores it. Only the status changes.
delete_planned_workoutwriteRemoves a planned day and returns it, so you can undo.
log_workoutwriteLogs a finished workout to the user's history. Full access.
update_workoutwriteChanges one finished workout at its current version: fields, exercises or sets. Full access.
delete_workoutwriteDeletes one finished workout at its current version. Full access.
restore_workoutwriteMakes an earlier version of a finished workout current again, within 7 days, also after a delete. Full access.
send_feedbackwriteSends a message from the user to the Tally team. Limit: 5 messages a day.

Rules

  • Finished workouts change only when the user asks. log_workout, update_workout, delete_workout and restore_workout need a Full access key or an OAuth connection. update_workout and delete_workout need the workout's version from get_workout as expectedVersion. Each change keeps the earlier content for 7 days: get_workout_versions lists it and restore_workout brings it back. All except restore_workout need a phone on the current Tally app.
  • Some days are protected. A day the user built in the app, or edited, is protected. The plan, update, move and delete tools skip it unless the user asks for an override.
  • Edits check the version. If a day or a finished workout changed after the agent read it, nothing is written and the result reports a conflict.
  • Each user sees only their own data. OAuth tokens and API keys are scoped to one account. The user can revoke them at any time.

Links

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.