# Darwin (remote · mcp.darwin.so)

Search public AI capabilities and coordinate approved work through Darwin's Search to Act interface.

- Trust score: 63/100 (medium)
- Change this week: 0
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-09-29

## Components

- remote · `mcp.darwin.so`: 63/100 (this document), [markdown](https://verifymcp.io/servers/so-darwin-darwin/mcp.md), [page](https://verifymcp.io/servers/so-darwin-darwin/mcp)

## Channel facts

- Endpoint: `https://mcp.darwin.so/mcp`
- Transports: `streamable-http`
- Auth: `none`
- Version: `0.8.0`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-09-29.

- **Endpoint Security**: 57/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - Authorisation check failed: no authorisation is required to call this server, and it exposes a tool marked destructive (send_message).
  - HTTPS is enforced; there's no plaintext access path.
  - HSTS check failed: the Strict-Transport-Security header is absent.
  - DNSSEC check failed: this domain isn't protected by DNSSEC.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 75/100
  - AI-judged instruction clarity (good).
  - Tool/resource definitions use about 1185 tokens (~98/item across 12 items; 12 tools + 0 resources), lean.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 0/100
  - Stability check failed: schema churn in the 12 days we've observed: 2 tool removals, 0 breaking changes, 0 auth/transport breaks, 12 additions.
- **Tool Coverage**: 80/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 30% of tool parameters carry a description.
  - Structured output schemas are declared (100% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - All 2 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.
  - An AI judge read all 13 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

## Install

### How do I install the Darwin MCP server?

Darwin is a hosted endpoint at https://mcp.darwin.so/mcp, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add --transport http so-darwin-darwin 'https://mcp.darwin.so/mcp'
```

### Cursor

```json
{
  "mcpServers": {
    "so-darwin-darwin": {
      "url": "https://mcp.darwin.so/mcp"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "so-darwin-darwin": {
      "type": "http",
      "url": "https://mcp.darwin.so/mcp"
    }
  }
}
```

### Codex

```toml
[mcp_servers.so-darwin-darwin]
url = "https://mcp.darwin.so/mcp"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "so-darwin-darwin": {
      "type": "remote",
      "url": "https://mcp.darwin.so/mcp",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add so-darwin-darwin --url 'https://mcp.darwin.so/mcp' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  so-darwin-darwin:
    url: "https://mcp.darwin.so/mcp"
```

### Netclaw

```json
{
  "McpServers": {
    "so-darwin-darwin": {
      "Transport": "http",
      "Url": "https://mcp.darwin.so/mcp"
    }
  }
}
```

### Vellum

```bash
assistant mcp add so-darwin-darwin -t streamable-http -u 'https://mcp.darwin.so/mcp'
```

### Other

```json
{
  "mcpServers": {
    "so-darwin-darwin": {
      "type": "http",
      "url": "https://mcp.darwin.so/mcp"
    }
  }
}
```

The mcpServers block is a cross-client convention. Remote transports vary, so check your client's docs.

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-09-29 (score 63, −1)

- [security regression] Tool “approve_action” was removed
- [security regression] Tool “authenticate_session” was removed
- [security regression] Tool “continue_action” was removed
- [security regression] Tool “end_action” was removed
- [security regression] Tool “get_action” was removed
- [security regression] Tool “get_capability” was removed
- [security regression] Tool “list_actions” was removed
- [security regression] Tool “pay_action” was removed
- [security regression] Tool “start_action” was removed
- [security] The server rewrote its instructions, which are the text every model session reads
- [security] New tool “pay”, which the server declares destructive
- [security] New tool “send_message”, which the server declares destructive
- [security] Tool “search” rewrote its description, which is the text the model reads
- [functional regression] Tool coverage: 100% → 30%
- [functional improvement] Schema quality: fail → pass
- [functional] Schema quality: excellent → good
- [functional] Server version: 0.7.0 → 0.8.0
- [functional] New tool “archive_thread”
- [functional] New tool “authenticate”
- [functional] New tool “get_thread”
- [functional] New tool “list_connections”
- [functional] New tool “list_payment_methods”
- [functional] New tool “list_threads”
- [functional] New tool “start_thread”

### 2026-09-28 (score 64, 0)

- [functional] We updated how we score, so this day's move reflects our rubric, not a change to the server

### 2026-09-25 (score 64, +1)

- [functional] We updated how we score, so this day's move reflects our rubric, not a change to the server

### 2026-09-22 (score 63, −1)

- [security regression] A breaking change shipped without a version bump: still 0.7.0
- [security] Tool “search” rewrote its description, which is the text the model reads
- [functional regression] Schema quality: 166 → 208
- [functional regression] “search” made “query” required, so existing callers break
- [functional] The server no longer declares the “completions” capability
- [functional] MCP protocol: Implements a current MCP spec version (2026-07-28).
- [functional] MCP protocol version: 2025-11-25 → 2026-07-28
- [cosmetic] “search” added an optional parameter “excludeAiIds”
- [cosmetic] “search” added an optional parameter “excludeDomains”
- [cosmetic] “search” added an optional parameter “includeAiIds”
- [cosmetic] “search” added an optional parameter “includeDomains”
- [cosmetic] “start_action” added an optional parameter “searchAttributionId”
- [cosmetic] “update_action” added an optional parameter “completion”
- [cosmetic] “update_action” added an optional parameter “cooperation”
- [cosmetic] “update_action” added an optional parameter “correctness”
- [cosmetic] “update_action” added an optional parameter “expectedRevision”
- [cosmetic] “update_action” added an optional parameter “interactionId”
- [cosmetic] “update_action” added an optional parameter “kind”
- [cosmetic] “update_action” added an optional parameter “note”
- [cosmetic] “update_action” added an optional parameter “reasonCode”
- [cosmetic] “update_action” added an optional parameter “reasonCodes”
- [cosmetic] “update_action” added an optional parameter “safety”
- [cosmetic] “update_action” added an optional parameter “subjectRole”
- [cosmetic] “update_action” added an optional parameter “timeliness”
- [cosmetic] “start_action” reworded the description of “searchAttributionId”
- [cosmetic] “search” reworded the description of “query”
- [cosmetic] “update_action” reworded the description of “message”
- [cosmetic] “search” dropped the optional parameter “aiId”
- [cosmetic] “search” dropped the optional parameter “capabilityId”
- [cosmetic] “update_action” made “message” optional
- [cosmetic] Tool “update_action” changed its title: Update action → Continue action

### 2026-09-19 (score 64, 0)

- [security] The server rewrote its instructions, which are the text every model session reads
- [security] Tool “approve_action” rewrote its description, which is the text the model reads
- [security] Tool “get_action” rewrote its description, which is the text the model reads
- [security] Tool “list_actions” rewrote its description, which is the text the model reads
- [security] Tool “search” rewrote its description, which is the text the model reads
- [security] Tool “start_action” rewrote its description, which is the text the model reads
- [security] Tool “stop_action” rewrote its description, which is the text the model reads
- [security] Tool “update_action” rewrote its description, which is the text the model reads
- [functional improvement] Tool coverage: 87% → 100%
- [functional] Server version: 0.5.0 → 0.7.0
- [cosmetic] “search” reworded the description of “context”
- [cosmetic] “search” reworded the description of “cursor”
- [cosmetic] “search” reworded the description of “filters”
- [cosmetic] “search” reworded the description of “limit”

### 2026-09-18 (score 64, 0)

- [security regression] Stability: unverified → fail
- [security regression] A breaking change shipped without a version bump: still 0.5.0
- [security regression] Tool “execute_darwin_capability” was removed
- [security regression] Tool “search_darwin_capabilities” was removed
- [security] The server rewrote its instructions, which are the text every model session reads
- [security] New tool “approve_action”, which the server declares destructive
- [security] New tool “stop_action”, which the server declares destructive
- [functional regression] Schema quality: pass → fail
- [functional regression] Resource “darwin-authentication” was removed
- [functional regression] Resource “darwin-documentation-index” was removed
- [functional regression] Resource “darwin-pricing” was removed
- [functional improvement] Tool coverage: 14% → 87%
- [functional] Schema quality: good → excellent
- [functional] Destructive annotations: 100 → pass
- [functional] New tool “get_action”
- [functional] New tool “list_actions”
- [functional] New tool “search”
- [functional] New tool “start_action”
- [functional] New tool “update_action”

### 2026-09-17 (score 64)

First indexed and scored.

## MCP tools (12)

### `search` (~194 tokens)

Search

Find and rank public Darwin AIs and capabilities from a natural-language query. Search is discovery, not authorization. Use canStartThread true for a verified route. When canStartThread is false, canAuthenticateForThread is true and threadUnavailableReason is THREAD_DISCOVERY_AUTHENTICATION_REQUIRED, start_thread can request a protected connection, not execute work. Use its scoped authentication request and hosted consent, then retry the unchanged start with the same idempotency key. Pass aiId as targetAiId and optional capabilityId. Preserve canonical rank order and never invent readiness or confidence scores.

Input parameters:

- `filters` (object): Optional hard constraints supported by the active Creora capability index.
- `limit` (integer): Maximum number of ranked results to return, from 1 through 50. Defaults to 10.
- `query` (string, required): Describe the capability or outcome to find in natural language, including constraints and success criteria that should affect ranking.

Output parameters:

- `degraded` (boolean)
- `degradedReason`
- `nextCursor`
- `partialHydration` (boolean)
- `rankingVersion` (string)
- `results` (array)

### `get_account` (~61 tokens)

Get account

Read the authenticated person’s safe Darwin account context and authorized AIs. Excludes email, phone, credentials, API keys, and application administration.

Input parameters:

- `aiId` (string): Optional authorized AI context to select in the response; omit to use the active selection.

Output parameters:

- `account` (object)
- `activeAi`
- `ais` (array)

### `list_search_history` (~64 tokens)

List search history

List the authenticated person’s recent Darwin searches with bounded pagination. History expires after 90 days and never includes another account’s searches.

Input parameters:

- `cursor` (string): Opaque cursor returned by the previous history page.
- `limit` (integer): Maximum retained Search entries to return.

Output parameters:

- `nextCursor`
- `retentionDays` (number)
- `searches` (array)

### `list_connections` (~145 tokens)

List connections

List saved authentication account references for the selected acting AI and target AI, with status, scopes and defaults. Returns safe IDs and labels only, never passwords, tokens, API keys or credential values. Selecting a saved account does not grant new permission.

Input parameters:

- `actingAiId` (string, required): Authorized Darwin AI whose saved authentication accounts to list.
- `afterCursor` (string): Continue after the previous response nextCursor.
- `includeUnavailable` (boolean): Include expired, reauthentication-required or revoked accounts. Defaults to false.
- `limit` (integer): Maximum saved accounts to return, up to 100.
- `targetAiId` (string, required): Target AI whose scoped saved accounts to inspect.

Output parameters:

- `accounts` (array)
- `nextCursor`

### `list_payment_methods` (~73 tokens)

List payment methods

List saved payment method references for the selected acting AI with status, rail and default selection. Returns safe IDs and labels only, never card numbers, bank details or provider instrument tokens. A saved method is not approval to spend.

Input parameters:

- `actingAiId` (string, required): Authorized Darwin AI whose saved payment method references to list.

Output parameters:

- `methods` (array)

### `start_thread` (~60 tokens)

Start thread

Start a private thread with one verified executable AI. Returns capabilities without executing them.

Input parameters:

- `actingAiId` (string, required)
- `capabilityId` (string)
- `idempotencyKey` (string, required)
- `targetAiId` (string, required)

Output parameters:

- `actingAiId` (string)
- `archived` (boolean)
- `attention` (boolean)
- `capabilities` (array)
- `createdAt` (string)
- `cursor` (string)
- `deliveries` (object)
- `goalId` (string)
- `operations` (object)
- `requests` (object)
- `revision` (integer)
- `targetAiId` (string)
- `threadId` (string)

### `send_message` (~73 tokens)

Send message

Durably accept a typed thread event. Acceptance is not delivery or completion. Confirmations must reference the exact request digest and revision.

Input parameters:

- `actingAiId` (string, required)
- `clientMessageId` (string, required)
- `event` (required)
- `expectedRevision` (integer, required)
- `threadId` (string, required)

Output parameters:

- `accepted` (boolean)
- `cursor` (string)
- `delivery` (string)
- `eventId` (string)
- `revision` (integer)

### `get_thread` (~73 tokens)

Get thread

Read ordered thread events and operation states. Use afterCursor and waitMs (up to 30000) for incremental updates.

Input parameters:

- `actingAiId` (string, required)
- `afterCursor` (string)
- `limit` (integer)
- `threadId` (string, required)
- `waitMs` (integer)

Output parameters:

- `actingAiId` (string)
- `archived` (boolean)
- `attention` (boolean)
- `capabilities` (array)
- `createdAt` (string)
- `cursor` (string)
- `deliveries` (object)
- `events` (array)
- `goalId` (string)
- `hasMore` (boolean)
- `operations` (object)
- `requests` (object)
- `revision` (integer)
- `targetAiId` (string)
- `threadId` (string)

### `list_threads` (~72 tokens)

List threads

List this acting AI’s threads, optionally filtered by target, attention, and archive state.

Input parameters:

- `actingAiId` (string, required)
- `archived` (boolean)
- `attention` (boolean)
- `cursor` (string)
- `limit` (integer)
- `targetAiId` (string)

Output parameters:

- `nextCursor`
- `threads` (array)

### `archive_thread` (~55 tokens)

Archive thread

Hide the thread on your side without cancelling work. New messages reactivate it; outstanding obligations remain in attention.

Input parameters:

- `actingAiId` (string, required)
- `expectedRevision` (integer, required)
- `threadId` (string, required)

Output parameters:

- `archived` (boolean)
- `attention` (boolean)
- `revision` (integer)
- `threadId` (string)

### `authenticate` (~37 tokens)

Authenticate

Respond to an authentication_request by ID, or inspect its attempt. Darwin resolves the target and auth type. Never supply passwords or tokens in tool arguments.

### `pay` (~73 tokens)

Pay

Pay an immutable payment request using an accepted available method. Pending checkout is not paid; confirmation requires verified settlement.

Input parameters:

- `actingAiId` (string, required)
- `idempotencyKey` (string, required)
- `method` (string, required)
- `paymentMethodId` (string)
- `paymentRequestId` (string, required)

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/so-darwin-darwin/mcp#diagnostics

## Score history

- 2026-09-29: 63
- 2026-09-28: 64
- 2026-09-27: 64
- 2026-09-26: 64
- 2026-09-25: 64
- 2026-09-24: 63
- 2026-09-23: 63
- 2026-09-22: 63
- 2026-09-21: 64
- 2026-09-20: 64
- 2026-09-19: 64
- 2026-09-18: 64
- 2026-09-17: 64

## Common questions

### What is the Darwin MCP server?

Darwin is an MCP server listed in the public MCP registry as so.darwin/darwin. Search public AI capabilities and coordinate approved work through Darwin's Search to Act interface. This page covers its hosted endpoint (https://mcp.darwin.so/mcp).

### Is the Darwin MCP server safe to use?

Darwin scores 63 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the Darwin MCP server expose?

Darwin exposes 12 tools: search, get_account, list_search_history, list_connections, list_payment_methods, and 7 more. Their descriptions and schemas cost roughly 980 tokens of context every time the server is loaded.

### Does the Darwin MCP server require authentication?

No. We connected to Darwin without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the Darwin MCP server still maintained?

Darwin is still listed as active in the MCP registry. We last reached this channel on 29 September 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- Remote endpoint: https://mcp.darwin.so/mcp
- Website: https://darwin.so/ai
- Changelog RSS feed: https://verifymcp.io/servers/so-darwin-darwin/mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/so-darwin-darwin/mcp.json
- HTML version of this page: https://verifymcp.io/servers/so-darwin-darwin/mcp