AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
MCP unreviewed MIT Self-run

Hermes IOS

mcp-dylan-buck-hermes-ios · by dylan-buck

Hermes agents don't have iPhones. Now yours does. This app gives Hermes camera, mic, health data, location, and notifications. All available in one MCP tool.

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add mcp-dylan-buck-hermes-ios

Open-source listing, not yet scanned by AgentStack. Follow the source repository for install instructions.

Security review

⚠ Flagged

1 finding(s); flagged for manual review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures
  • high Destructive filesystem operation.

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets Used
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Reliability & compatibility

Not yet reviewed
0 installs to date
no reviews yet
3mo ago

Declared compatibility

Claude CodeClaude DesktopCursorWindsurf

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Hermes IOS? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Hermes iOS

> [!NOTE] > Hermes iOS is an independent community project. It is not affiliated with, endorsed by, or part of Nous Research or the official Hermes Agent project.

Hermes iOS is a self-hosted-first iPhone companion for a user-owned Hermes runtime. It adds a native iOS app, a public relay, and a host-side connector so Hermes can move between desktop, phone, sensors, widgets, and voice without turning your runtime into a hosted service.

Why use it

  • Self-hosted first: the relay is yours, the connector runs next to your Hermes install, and the iPhone app can point to any compatible relay.
  • Native iPhone experience: chat, voice mode, camera attachments, widgets, Live Activities, and sensor-aware context.
  • Hermes-aware: slash commands, installed skills, personalities, quick commands, MCP-backed context, and agent-side coding workflows.
  • Optional platform extras: APNs and CarPlay are supported, but not required to get a working setup.

What works today

  • Streaming chat with retries, attachments, inline diffs, markdown/code blocks, and inline returned images
  • Voice mode with OpenAI Realtime, live camera context, and Hermes tool delegation
  • Dynamic slash-command catalog sourced from Hermes built-ins, installed skills, personalities, and quick-command metadata
  • Health, location, motion, and sensor storage through the connector SQLite pipeline
  • Home Screen widgets, Live Activities, and host/model/context status in the chat UI
  • Self-hosted relay + connector pairing flow with background service support

Sensor Pipeline & MCP Tools

The connector continuously receives sensor data from your iPhone and stores it locally in SQLite on your Hermes machine. Hermes accesses this data through MCP tools — meaning your agent can answer questions about you, build on top of your personal data, and use it as context in any conversation.

What flows from your phone:

  • Real-time GPS location with reverse-geocoded addresses
  • 11 HealthKit metrics (steps, heart rate, sleep, calories, blood oxygen, and more)
  • CoreMotion activity detection (walking, running, driving, cycling, stationary)

What your agent can do with it:

You:    "Find me a good dinner spot tonight"
Hermes: Looks like you've been around the West Village today. Carbone, Via
        Carota, and L'Artusi are all nearby and well-reviewed. Carbone usually
        needs a reservation — want me to look into availability?

You:    "Am I on track for my fitness goals this week?"
Hermes: You've hit 42,000 steps so far this week (avg 8,400/day), burned
        2,100 active calories, and averaged 7.1 hours of sleep. Your
        resting heart rate has been steady at 62 bpm. You're ahead on
        steps but slightly behind on sleep — you lost about an hour
        last night compared to your weekly average.

You:    "Build me a personal health dashboard"
Hermes: I'll query your sensor database and create an interactive dashboard
        with your steps, sleep trends, heart rate, and location history
        over the past 30 days. Give me a minute.

The data never leaves your machine. The sensor SQLite database lives at ~/.hermes-mobile/sensors.db and is queryable through the hermes_mobile MCP server with 8 tools:

| Tool | What it does | |------|-------------| | get_user_location | Current location with address | | get_location_history | Where you've been (time-series) | | get_health_summary | All latest health metrics at once | | get_health_metric | Time-series for one metric | | get_health_metrics_list | Available metrics + latest values | | get_user_activity | Current physical activity | | get_sensor_schema | Database table definitions | | query_sensor_data | Custom read-only SQL queries |

The query_sensor_data tool is particularly powerful — your agent can write arbitrary SELECT queries against your health and location data to find correlations, build charts, generate reports, or power any project that benefits from personal sensor data.

> [!TIP] > Install the bundled hermes-ios skill (see step 5 below) to teach Hermes when and how to use these tools effectively.

Architecture

flowchart LR
    A["iOS app"] -->|"HTTP / SSE"| B["Relay"]
    B -->|"WebSocket jobs / RPC"| C["Connector"]
    C -->|"CLI / API runtime"| D["Hermes Agent"]
    A -->|"sensor uploads"| B
    C -->|"SQLite + MCP tools"| D

The relay is the control plane. In connector mode, Hermes execution stays on the user-owned machine where the connector is installed.

> [!IMPORTANT] > If you test on a physical iPhone against a local relay, use your Mac's LAN IP or a public URL. Do not use 127.0.0.1 or localhost; on a phone those point back to the phone itself.

Quick start

1. Run the relay

cd relay
python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
cp .env.example .env
uvicorn app.main:app --reload

For a production-style setup, see [relay/README.md](relay/README.md) and [relay/docs/fly-io.md](relay/docs/fly-io.md).

2. Install and set up the connector

cd connector
python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]

export HERMES_COMMAND=/absolute/path/to/hermes
export HERMES_MOBILE_RELAY_URL=https://your-relay.example.com/v1

hermes-mobile setup

The setup wizard can:

  • use an existing relay URL
  • guide you through a Fly.io relay deployment
  • register the local hermes_mobile MCP server
  • configure OpenAI Realtime talk mode
  • install the connector background service

See [connector/README.md](connector/README.md) for the full flow.

3. Build and open the iPhone app

  1. Open the project in Xcode.
  2. Set your signing team and local bundle/App Group overrides if needed.
  3. Launch the app on your phone or simulator.
  4. Enter the same relay URL used by the connector.

Connect Hermes screen with a custom relay URL entered and the phone pairing code ready to scan.

4. Pair the phone

On the connector host:

hermes-mobile pair-phone

Then scan the QR code or enter the manual code in the iPhone app.

5. Recommended: install the Hermes iOS skill in Hermes

The connector exposes the hermes_mobile MCP tools, but the bundled hermes-ios skill teaches Hermes when and how to use them well for location, health, activity, and sensor-aware responses.

From the repo root:

mkdir -p ~/.hermes/skills
cp -R skills/hermes-ios ~/.hermes/skills/

If you are updating an older local copy of the skill, replace it explicitly:

rm -rf ~/.hermes/skills/hermes-ios
cp -R skills/hermes-ios ~/.hermes/skills/

Then reload Hermes:

/reload-mcp

Or start a fresh Hermes chat/session if you prefer.

Optional features

> [!TIP] > You do not need APNs or CarPlay to get a working setup. They are additive features.

  • APNs: lets the relay deliver Hermes replies while the app is backgrounded. Setup lives in [docs/CONFIGURATION.md](docs/CONFIGURATION.md).
  • CarPlay: requires Apple approval for the voice-based conversational entitlement. It is optional and inert when not configured.

Documentation map

  • [connector/README.md](connector/README.md): connector install, wizard flow, pairing, service management, and troubleshooting
  • [relay/README.md](relay/README.md): relay setup, production checklist, API surface, and deployment notes
  • [docs/CONFIGURATION.md](docs/CONFIGURATION.md): environment variables, iOS build settings, APNs, CarPlay, and private overrides
  • [relay/docs/fly-io.md](relay/docs/fly-io.md): manual Fly.io deployment steps
  • [relay/docs/local-dev.md](relay/docs/local-dev.md): local development and same-network testing notes
  • [connector/SENSORSCHEMA.md](connector/SENSORSCHEMA.md): connector SQLite schema and MCP query surface
  • [docs/IOSCAPABILITIES.md](docs/IOSCAPABILITIES.md): technical iOS capability ledger for maintainers
  • [MAINTAINERNOTES.md](MAINTAINERNOTES.md): maintainer-facing implementation snapshot, not onboarding

Project status

Hermes iOS is already usable as a self-hosted project, but it is still actively evolving. The public docs are optimized for getting a new user from clone to first paired chat as quickly as possible; the maintainer docs track deeper capability and architecture details.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.