> ## Documentation Index
> Fetch the complete documentation index at: https://cyberpaisa-dof-mesh-40-27.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Config Reference

> mcp_config.py — MCP_REGISTRY, ROLE_MCP_MAP, get_mcp_for_role(), and 11 available servers.

`mcp_config.py` is the single source of truth for MCP server configuration.
It uses a two-table architecture: `MCP_REGISTRY` defines every available server,
and `ROLE_MCP_MAP` assigns servers to agent roles — no hardcoded logic.

```
MCP_REGISTRY   → defines each server (command, args, env)
ROLE_MCP_MAP   → assigns MCPs to agent roles
get_mcp_for_role() → builds MCPServerStdio objects on demand
```

***

## MCP\_REGISTRY — 11 Servers

### Foundational (4)

| Key          | Package                                   | Description                                       |
| ------------ | ----------------------------------------- | ------------------------------------------------- |
| `filesystem` | `@modelcontextprotocol/server-filesystem` | Read/write files in allowed directories           |
| `web_search` | `@pskill9/web-search`                     | Google search — no API key required               |
| `fetch`      | `@anthropics/mcp-server-fetch`            | Fetch URLs and convert to markdown                |
| `memory`     | `@anthropics/mcp-server-memory`           | Persistent knowledge graph (complements ChromaDB) |

### Guardian-audited (7 — 26 March 2026)

| Key                   | Package                                            | Publisher                            | API key                 |
| --------------------- | -------------------------------------------------- | ------------------------------------ | ----------------------- |
| `context7`            | `@upstash/context7-mcp`                            | Upstash (official) — 430K dl/week    | —                       |
| `sequential_thinking` | `@modelcontextprotocol/server-sequential-thinking` | Anthropic (official) — 94K dl/week   | —                       |
| `playwright`          | `@playwright/mcp --headless`                       | Microsoft (official) — 1.86M dl/week | —                       |
| `evm`                 | `@mcpdotdirect/evm-mcp-server`                     | mcpdotdirect — SLSA provenance       | —                       |
| `tavily`              | `tavily-mcp`                                       | Tavily AI (official) — 137K dl/week  | `TAVILY_API_KEY`        |
| `brave_search`        | `@brave/brave-search-mcp-server`                   | Brave Software (official)            | `BRAVE_API_KEY`         |
| `supabase`            | `@supabase/mcp-server-supabase`                    | Supabase (official) — 47K dl/week    | `SUPABASE_ACCESS_TOKEN` |

<Warning>
  `evm` — **NEVER** connect your primary wallet. Use a dedicated wallet only.
</Warning>

***

## ROLE\_MCP\_MAP

### Crew roles (8)

| Role                | MCPs                                          |
| ------------------- | --------------------------------------------- |
| `code_architect`    | filesystem, fetch, context7, evm, playwright  |
| `research_analyst`  | tavily, brave\_search, context7, fetch        |
| `data_engineer`     | filesystem, supabase, context7                |
| `project_organizer` | filesystem, supabase                          |
| `narrative_content` | brave\_search, fetch, context7                |
| `mvp_strategist`    | sequential\_thinking, brave\_search, context7 |
| `qa_reviewer`       | playwright, context7, tavily                  |
| `verifier`          | evm, sequential\_thinking, tavily             |

### Agent roles (17 — agents.yaml)

| Agent ID                | MCPs                                             |
| ----------------------- | ------------------------------------------------ |
| `lider_orquestador`     | sequential\_thinking, tavily, context7, supabase |
| `organizador_archivos`  | filesystem                                       |
| `product_manager`       | sequential\_thinking, brave\_search, context7    |
| `director_operaciones`  | supabase, tavily, sequential\_thinking           |
| `bizdev`                | brave\_search, tavily, fetch                     |
| `arquitecto`            | filesystem, context7, evm, sequential\_thinking  |
| `developer`             | filesystem, context7, fetch, playwright          |
| `qa_engineer`           | playwright, context7, tavily                     |
| `investigador`          | tavily, brave\_search, fetch, context7           |
| `devops`                | filesystem, supabase, playwright                 |
| `seguridad_blockchain`  | evm, sequential\_thinking, tavily                |
| `experto_ideacion`      | brave\_search, fetch, sequential\_thinking       |
| `experto_multichain`    | evm, tavily, context7                            |
| `experto_cuantico`      | sequential\_thinking, context7, tavily           |
| `seguridad_informatica` | evm, playwright, sequential\_thinking            |
| `experto_metodologias`  | sequential\_thinking, context7, fetch            |
| `experto_bpm`           | sequential\_thinking, supabase, context7         |

***

## API

### get\_mcp\_for\_role()

Returns a list of `MCPServerStdio` objects for a given agent role.
Accepts both Spanish (`agents.yaml`) and English alias names.

```python theme={null}
from mcp_config import get_mcp_for_role

# By role name (Spanish)
mcps = get_mcp_for_role("investigador")
# → [MCPServerStdio(tavily), MCPServerStdio(brave_search), ...]

# By alias (English)
mcps = get_mcp_for_role("researcher")
# → same as "investigador"

# With custom project directory (for filesystem MCP)
mcps = get_mcp_for_role("developer", project_dir="/tmp/my-project")
```

### get\_available\_mcps()

```python theme={null}
from mcp_config import get_available_mcps

registry = get_available_mcps()
# {
#   "filesystem": {"description": "...", "env_required": None, "warning": None},
#   "evm":        {"description": "...", "env_required": None,
#                  "warning": "NUNCA conectar wallet principal..."},
#   "tavily":     {"description": "...", "env_required": "TAVILY_API_KEY", ...},
#   ...
# }
```

### get\_role\_map()

```python theme={null}
from mcp_config import get_role_map

role_map = get_role_map()
print(role_map["code_architect"])
# ["filesystem", "fetch", "context7", "evm", "playwright"]
```

***

## Adding a New MCP Server

1. Add an entry to `MCP_REGISTRY`:

```python theme={null}
"my_server": {
    "command": "npx",
    "args": ["-y", "@org/my-mcp-server"],
    "description": "What it does",
    "env": "MY_API_KEY",  # optional
},
```

2. Add to relevant roles in `ROLE_MCP_MAP`:

```python theme={null}
"my_role": ["filesystem", "my_server"],
```

No other changes needed — `get_mcp_for_role()` picks it up automatically.

***

## Role Aliases (English → Spanish)

| English alias         | Maps to                 |
| --------------------- | ----------------------- |
| `orchestrator`        | `lider_orquestador`     |
| `file_organizer`      | `organizador_archivos`  |
| `operations`          | `director_operaciones`  |
| `architect`           | `arquitecto`            |
| `security_blockchain` | `seguridad_blockchain`  |
| `security`            | `seguridad_informatica` |
| `researcher`          | `investigador`          |
| `ideation`            | `experto_ideacion`      |
| `multichain`          | `experto_multichain`    |
| `quantum`             | `experto_cuantico`      |
| `methodologies`       | `experto_metodologias`  |
| `bpm`                 | `experto_bpm`           |

***

<CardGroup cols={2}>
  <Card title="MCP & API Guide" icon="plug" href="/guides/mcp-api">
    REST API endpoints and DOF MCP tools
  </Card>

  <Card title="Providers & Agents" icon="robot" href="/guides/providers-and-agents">
    17 agents and their roles
  </Card>
</CardGroup>