mcp-server-hevy
Enables access to Hevy strength-training data—workouts, routines, exercise history, and progression—via an MCP server with per-user credential encryption and optional write support.
README
mcp-server-hevy
MCP server for the Hevy strength-training API — workouts, routines, exercise history and progression data.
Self-hosted alternative to the public hosted Hevy MCP, built to the same shape as our other personal connectors: one shared instance, per-user credentials encrypted at rest, access gated by the identity the gateway verifies.
Independent, unofficial project. Not affiliated with Hevy.
Requirements
Hevy Pro. The Hevy developer API is only available to Pro subscribers — the API key simply doesn't exist on a free account, so nothing here will work without one.
How a user links their account
The key never passes through the conversation. Instead:
- Ask Claude to connect Hevy →
hevy_connectreturns a one-time link - Open it, paste the Hevy API key into the form, submit
- The server validates it against Hevy before storing it, so a wrong key fails immediately rather than as a confusing 401 later
The key is found at hevy.com/settings?developer — in a browser, not the mobile app.
This mirrors mcp-server-garmin: the enrollment route is reached directly
(bypassing the gateway) and is secured by a single-use, user-bound state token.
Unlike an OAuth code, a Hevy API key is long-lived, which is exactly why it
shouldn't sit in a chat transcript.
Tools
Auth: hevy_connect, hevy_status, hevy_disconnect, hevy_search_capabilities
Read: hevy_get_workouts, hevy_get_workout, hevy_get_workout_count,
hevy_get_workout_events (incremental sync), hevy_get_exercise_history,
hevy_get_routines, hevy_get_routine, hevy_get_routine_folders,
hevy_search_exercise_templates, hevy_get_exercise_template,
hevy_get_user_info, hevy_get_body_measurements
Write (only registered when HEVY_ENABLE_WRITES=true):
hevy_create_workout, hevy_update_workout, hevy_create_routine,
hevy_update_routine, hevy_create_routine_folder,
hevy_create_exercise_template
Hevy API constraints worth knowing
These are properties of Hevy's API, not choices made here:
- No DELETE, no PATCH. Nothing can be deleted through the API, and there is no partial update.
- PUT replaces the whole record. Any exercise or set omitted from an update
is erased. Always read the current object first and resend it complete — the
update tools are annotated
destructiveHintfor this reason. - Creates are not idempotent. A retried create makes a duplicate, so this client retries GET only; failed writes surface to the caller instead of silently double-posting.
- Units are fixed: weights in kilograms, durations in seconds.
- Small pages. Max 10 per page for workouts/routines/folders, 100 for exercise templates.
- Hevy's own docs describe the API as unstable and subject to change.
Body measurements are read-only, deliberately
Hevy can store body weight and composition, and its API can write them — this server intentionally doesn't. Withings is the source of truth for body metrics in our setup; writing them here too would create a second competing record with no defined winner. Reading is exposed so existing Hevy data stays visible.
Configuration
HEVY_ENCRYPTION_KEY= # AES-256-GCM key for the credential store (openssl rand -hex 32)
HEVY_STORE_PATH=/data/store.json
HEVY_PUBLIC_BASE_URL= # public origin, used to build the enrollment link
HEVY_TRUST_FORWARDED_USER=true
HEVY_ENABLE_WRITES=true
HEVY_AUDIT_LOG=
HEVY_TIMEOUT_MS=30000
MCP_HTTP_TOKEN=
No app-level API credentials exist — every credential is per user.
Verification
npm run typecheck && npm test && npm run build
License
Apache-2.0
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。