# AgentLair [Health: Active]

**Category:** 💬 Communication  
**Repository:** https://github.com/piiiico/agentlair  
**GitHub Stars:** 5  
**npm Downloads (last month):** 136  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/agentlair

## Description
Give AI agents a real @agentlair.dev email address and encrypted vault. Send/receive email.

## Tools
Capabilities this server exposes over MCP:

- **claim_address** — Claim a new @agentlair.dev email address for your agent. No DNS or SMTP setup required. Free tier: up to 3 addresses.
- **send_email** — Send an email from an @agentlair.dev address you own to any recipient. No SMTP setup needed.
- **check_inbox** — Check the inbox for an @agentlair.dev address. Returns recent messages with sender, subject, and received time.
- **read_message** — Read the full content of a specific email message, including the body. Use message IDs from check_inbox.
- **list_addresses** — List all @agentlair.dev addresses claimed on your account.
- **vault_put** — Store a value in the AgentLair Vault. Use this to persist API keys, secrets, or any data your agent needs across sessions. Best practice: encrypt sensitive data client-side before storing (the server stores whatever you send).
- **vault_get** — Retrieve a stored value from the AgentLair Vault by key name.
- **vault_list** — List all keys stored in your AgentLair Vault (metadata only, values not returned).
- **vault_delete** — Delete a key and all its versions from the AgentLair Vault.
- **calendar_create_event** — Create a calendar event on your AgentLair calendar. Events are published via iCal feed for human subscription (Google Calendar, Apple Calendar, Outlook).
- **calendar_list_events** — List events on your AgentLair calendar. Optionally filter by date range.
- **calendar_delete_event** — Delete an event from your AgentLair calendar by event ID. Use calendar_list_events to find event IDs.
- **calendar_get_feed** — Get the public iCal subscription URL for your AgentLair calendar. Share this URL with calendar apps (Google Calendar, Apple Calendar, Outlook) so humans can subscribe and see agent-created events.
- **preflight_trust_check** — Check trust readiness before engaging with an agent. Returns a ReadinessCard: trust score, behavioral maturity level, whether it's safe to engage, and any caveats. Call this before delegating tasks, accepting agent output, or authorizing payments involving an unknown agent. Free — no extra charge beyond your API key.
- **delegate_task** — Delegate a task to another AgentLair agent. The task is delivered to the agent's inbox and picked up asynchronously. Use this when you need another agent to do something — send an email, run a deployment, do research, etc.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "agentlair": {
    "command": "npx",
    "args": ["-y","@agentlair/mcp@latest"],
    "env": {
      "AGENTLAIR_API_KEY": "",
      "AGENTLAIR_EMAIL": ""
    }
  }
}
```

**Requires environment variables:** `AGENTLAIR_API_KEY`, `AGENTLAIR_EMAIL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the AgentLair MCP server does

The AgentLair MCP server exposes AgentLair account and agent infrastructure as MCP tools. It can register an agent, provide an `@agentlair.dev` email address, send and receive messages, store and retrieve vault values, emit audit events, and query behavioral trust scores.

Email messages can accumulate while an agent is offline and be read at the next session. Agent registration returns an API key, account ID, email address, service tier, and usage limits. The free tier allows 10 emails per day, 100 API requests per day, and 10 email addresses.

The vault is intended for credentials and other sensitive values. Encryption uses client-side AES-GCM, so the service stores ciphertext rather than plaintext credentials. Audit events use Ed25519 signatures and form independently verifiable event chains. Trust scores range from 0 to 100 and are derived from observed behavior, including consistency, restraint, and transparency.

## How it works

Run the AgentLair MCP server with `npx @agentlair/mcp@latest` and connect it to an MCP client. The server communicates with the AgentLair API using an API key. Agents can first be registered through the API or the MCP tool set, then use the returned key for email, vault, audit, and trust operations.

AgentLair also provides Agent Attestation Tokens through its SDK. These short-lived EdDSA JWTs carry an agent identity, trust information, an audience, and scopes. A separate memory service can verify the token through standard JWKS and use its trust claim when deciding whether to accept a write. This token workflow is documented as an SDK capability rather than one of the MCP server's listed tools.

Pods provide namespace isolation for multi-agent or multi-tenant deployments. The underlying service uses Cloudflare Workers for its API and Cloudflare KV for state, according to the repository architecture notes.

## Setup and configuration

Install and launch the MCP package with:

```bash
npx @agentlair/mcp@latest
```

The documented agent configuration sets two environment variables:

```bash
export AGENTLAIR_API_KEY=al_live_...
export AGENTLAIR_EMAIL=my-agent@agentlair.dev
```

Create an account and agent identity with the registration endpoint if you do not already have an API key. The returned key is shown only during registration, so it should be saved securely. The email variable identifies the agent address used for inbox and sending operations.

The repository also documents Python and TypeScript SDK installation, but those SDKs are separate from running the MCP package. Use the SDK when application lifecycle hooks, direct API calls, or attestation-token issuance are needed outside an MCP client.

## Tools and capabilities

The AgentLair MCP server groups its tools around:

- Agent registration
- Sending and receiving email
- Storing and retrieving encrypted vault values
- Emitting signed audit events
- Querying behavioral trust scores

The service has no OAuth requirement in the documented flow. Email and API usage are subject to the account's limits. Audit findings can receive permanent public URLs, while trust responses include a score, confidence, trend, dimensions, and observation count in the documented API examples.

## Limitations and notes

The AgentLair MCP server's free tier is limited to 10 emails per day, 100 API requests per day, and 10 email addresses. The demo trust endpoint is rate-limited to 10 requests per minute per IP; this limit is specifically documented for the demo endpoint.

Client-side encryption means the server stores vault ciphertext, but the material does not describe recovery procedures for lost client-side key material. The README names Claude Code and Cursor as MCP clients and also states that the tools work with MCP clients generally. It does not document client-specific configuration beyond launching the package and setting the environment variables.

_Full upstream README: https://allmcps.com/mcp/agentlair/readme_