narwhalCAUTION
A local-first, privacy-first AI agent. Your conversations, projects, and work context stay on your machine.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
English · 简体中文
Narwhal
A local-first, privacy-first DeepSeek Harness desktop app.
An Electron-based desktop shell that runs a compatible DeepSeek Harness runtime on loopback only. Manage your own AI agent workbench, MCP servers, Cordis plugins (runtime extensions), and skills — all locally, with zero cloud lock-in.
⚠️ Early Release — Narwhal is a personal open-source project. This is the first public release: core features work, but there are known gaps (Windows/Linux builds, auto-update, advanced model presets, rich trajectory renderer, etc.). Bug reports, feature requests, and PRs are all welcome. Roadmap is tracked in GitHub Issues.
Screenshots
Main Workbench
Workspace picker, agent workbench, conversation history, and composer controls — all local, all loopback.
Integrations Marketplace
Narwhal's standout feature: a built-in marketplace for M
43ea1fb56497OBSERVED · 2026-10-12Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add narwhal --env APPLE_API_KEY=${APPLE_API_KEY} --env APPLE_API_KEY_ID=${APPLE_API_KEY_ID} --env APPLE_APP_SPECIFIC_PASSWORD=${APPLE_APP_SPECIFIC_PASSWORD} -- npx -y [email protected]{
"mcpServers": {
"narwhal": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"APPLE_API_KEY": "${APPLE_API_KEY}",
"APPLE_API_KEY_ID": "${APPLE_API_KEY_ID}",
"APPLE_APP_SPECIFIC_PASSWORD": "${APPLE_APP_SPECIFIC_PASSWORD}"
}
}
}
}Exposed tools (7)
6 read · 1 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
compact | read | Compact older conversation history |
export | read | Download this Session log as a ZIP archive |
feedback | read | Record feedback about this session |
goal | write | Set or view the goal for a long-running task |
permission | read | Switch the permission preset (sandbox mode + approval policy) |
plan | read | Enter or leave plan mode |
repo | read | Show git branch and recent commits (skill) |
Trust audit
CAUTIONgrade B · trust 89/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (5 observation(s))
- Network
- declared (5 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (6)
icon.icns
const mockAgent: AgentSnapshot = { state: 'ready', origin: 'http://127.0.0.1:18789' }import type { ChatItem } from '../../shared/desktop-contract'outDir: '../../dist/renderer',
highlight.js, react-markdown, rehype-highlight, remark-gfm, ws, yaml, zod, @electron/notarize
5. **No raw configuration exposure**: The renderer cannot read or receive raw host configuration, API keys, or runtime settings.
Gates applied: no_behavioural_pass.
43ea1fb56497full audit observations/trust-audit/mcp-server/yumingtao__narwhal.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-12 | 43ea1fb56497 | CAUTION | B | 89 | first audit |
Questions
What is the narwhal MCP server?
A local-first, privacy-first AI agent. Your conversations, projects, and work context stay on your machine.
What tools does narwhal expose?
7 in total: 6 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is narwhal safe to connect to an agent?
With care. The audit graded it B (89/100) and found 6 things worth knowing before you trust this server, listed below with the exact line each was found on.
What credentials does narwhal need?
It reads APPLE_API_KEY, APPLE_API_KEY_ID and APPLE_APP_SPECIFIC_PASSWORD from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does narwhal run?
It speaks streamable-http, so it runs as a service you connect to over the network. It is published on npm as narwhal at 0.2.0.
How current is this page?
The grade is for one exact copy of the source (43ea1fb56497), read on 2026-10-12. The repository is watched and re-audited when it changes.