> ## Documentation Index
> Fetch the complete documentation index at: https://latch.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents

> Manage the AI clients that connect to Latch.

An **agent** represents an AI client (like Claude Desktop or Cursor) that connects to Latch. Agents authenticate using API keys and can be tracked individually in the audit log.

## What is an Agent?

When an AI assistant connects through the Latch CLI, it authenticates as an agent. This lets you:

* **Track activity** per client in the audit log
* **See when agents are active** (last seen timestamp)
* **Revoke access** by deleting the agent or its API key

Multiple agents can connect to the same Latch instance:

* **Claude Desktop** → Latch → MCP Servers
* **Cursor** → Latch → MCP Servers
* **Your custom agent** → Latch → MCP Servers

Each agent authenticates with its own API key.

## API Keys

API keys authenticate agent connections to Latch. Each key is tied to a workspace.

### Creating an API Key

1. Go to **Settings** → **API Keys**
2. Click **Create API Key**
3. Give it a descriptive name (e.g., "MacBook Claude", "Work Cursor")
4. Copy the key immediately — it won't be shown again

<Warning>
  Keep your API keys secret. Anyone with the key can route traffic through your Latch workspace.
</Warning>

### Using API Keys

Pass the API key to the Latch CLI:

```bash theme={null}
npx @latchagent/cli@latest run \
  --api-key "latch_YOUR_KEY" \
  --upstream "my-upstream" \
  --upstream-command "npx" \
  --upstream-args "-y,@modelcontextprotocol/server-filesystem,/tmp"
```

Or in Claude Desktop config:

```json theme={null}
{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": [
        "@latchagent/cli@latest",
        "run",
        "--api-key", "latch_YOUR_KEY",
        "--upstream", "my-upstream",
        "--upstream-command", "npx",
        "--upstream-args", "-y,@modelcontextprotocol/server-filesystem,/tmp"
      ]
    }
  }
}
```

## Managing Agents

### Viewing Agents

Go to **Settings** → **Agents** to see:

* All registered agents
* When they were created
* When they were last active

### Revoking Access

To revoke an agent's access:

1. Go to **Settings** → **API Keys**
2. Delete the API key associated with that agent

The agent will no longer be able to authenticate.

## Best Practices

### Use Separate Keys for Each Client

Create different API keys for each machine or use case:

* `macbook-claude` — Your personal laptop
* `work-cursor` — Work development environment
* `ci-agent` — Automated testing

This lets you revoke access to one client without affecting others.

### Descriptive Names

Name your keys descriptively so you know what to revoke if something goes wrong:

* ✅ `christian-macbook-pro-2026`
* ✅ `staging-server-cursor`
* ❌ `key1`
* ❌ `test`

### Rotate Keys Periodically

For sensitive environments, rotate API keys periodically:

1. Create a new key
2. Update your client configs
3. Delete the old key

## Tracking Agent Activity

The [Audit Log](/audit-log) shows which agent made each request:

1. Go to **Audit Log**
2. Filter by agent to see activity from a specific client
3. Review patterns — unusual activity might indicate a compromised key

## Troubleshooting

### "Invalid agent key"

* Verify you copied the full key (starts with `latch_`)
* Check the key belongs to the correct workspace
* Ensure the key hasn't been deleted

### Agent not appearing in dashboard

* The agent registers on first connection
* Check that your CLI command is using the correct `--api-key`
* Verify the Latch server is reachable