Apple Health for AI agents: Claude Code, OpenClaw, Hermes and Codex

Which agent hosts can use Syncky?

Syncky Health sends the Apple Health types you choose from your iPhone to a Syncky agent on your own computer or server. Your AI agent reads them through the agent's local MCP server or its shell commands, and can write entries back to Apple Health for the types you allow on the iPhone. Free during launch; plans are coming later (pricing). No Syncky account required.

Host Setup route Reads Writes to Apple Health Guide
Claude Code Local MCP: claude mcp add --transport stdio syncky -- npx -y @syncky/agent mcp, or the Claude Code plugin The types you chose on the iPhone Types you allow for writing. The agent is instructed to send a write only with your explicit approval for that write or under a standing rule you approved; the iPhone applies writes only for allowed types and shows no prompt per write Claude Code
OpenClaw ClawHub skill: openclaw skills install @syncky/syncky, pinned to the agent release The types you chose on the iPhone Types you allow for writing. The agent is instructed to send a write only with your explicit approval for that write or under a standing rule you approved; the iPhone applies writes only for allowed types and shows no prompt per write OpenClaw
Hermes Give your Hermes agent https://syncky.app/agent (tested on Hermes), or install the shell skill by hand The types you chose on the iPhone Types you allow for writing. The agent is instructed to send a write only with your explicit approval for that write or under a standing rule you approved; the iPhone applies writes only for allowed types and shows no prompt per write Hermes and Apple Health
Codex CLI and IDE Local MCP in Codex's TOML configuration The types you chose on the iPhone Off in the guide's read-only configuration (syncky_send and syncky_ack disabled); enable deliberately Codex

Every route uses the same paired agent: pairing, receiving and permissions are described in agent setup. Other local clients have their own guides.

How does delivery work?

Delivery is automatic in the background whenever iOS lets the app run; it is not real time. Health reads require an unlocked iPhone, so new samples can arrive minutes to hours after they are recorded. Writes apply at the next sync, only for supported types you allowed, and each entry's HealthKit metadata records the requesting agent's IDs; the HealthKit writing app is Syncky Health.

The agent keeps what it received as raw JSON for 30 days. The 30-day inbox is a local buffer, not a limit on your Health history: Apple Health keeps your data on the iPhone, and the agent can request history again from the iPhone, within the history period the app sends; that does not guarantee recovering every sample.

What does Syncky see?

Client health-data envelopes are encrypted end to end between the iPhone and your agent. Relay sees delivery metadata and its own plaintext hub notices, but no health values, and queues sealed envelopes until acknowledgment or expiry, at most 7 days. Your agent receives plaintext locally, and your chosen model sees what the agent supplies.

Where do I compare other options?

See Apple Health MCP: options compared for export, cloud and local projects with dated checks, and writing to Apple Health for write permissions and source metadata.