{"_id":"@cognitiondesk/mcp","name":"@cognitiondesk/mcp","dist-tags":{"latest":"0.5.0"},"versions":{"0.5.0":{"name":"@cognitiondesk/mcp","version":"0.5.0","publishConfig":{"access":"public","registry":"https://registry.npmjs.org/"},"repository":{"type":"git","url":"git+https://github.com/Mounaji-Webs/mcp-cognitiondesk.git"},"homepage":"https://app.cognitiondesk.com","license":"MIT","keywords":["mcp","modelcontextprotocol","cognitiondesk","ai-assistants","workflows","claude"],"type":"module","description":"MCP server for the CognitionDesk platform: create and manage AI assistants, workflows, squads and knowledge, and inspect your usage — authenticated with your personal API key.","bin":{"mcp-cognitiondesk":"dist/index.js"},"scripts":{"dev":"tsx watch src/index.ts","start":"tsx src/index.ts","build":"tsc","typecheck":"tsc --noEmit","prepublishOnly":"tsc"},"dependencies":{"@modelcontextprotocol/sdk":"^1.25.2","dotenv":"^17.4.2","zod":"^3.24.0"},"//devDependencies":"Los @mounaji_npm/* son la capa de flota INTERNA (ver src/fleet.ts). Van acá y no en dependencies a propósito: son privados del scope cerrado, y publicarlos como dependencia haría que `npx @cognitiondesk/mcp` fallara con 404 en la máquina de cualquiera. Adentro se instalan desde vendor/ y todo funciona igual.","devDependencies":{"@mounaji_npm/agentfs":"file:vendor/mounaji_npm-agentfs-0.2.0.tgz","@mounaji_npm/mcp-agent":"file:vendor/mounaji_npm-mcp-agent-0.3.0.tgz","@mounaji_npm/mcp-telemetry":"file:vendor/mounaji_npm-mcp-telemetry-0.4.0.tgz","@types/node":"^22.0.0","tsx":"^4.19.0","typescript":"^5.7.0"},"gitHead":"6fca36ea3cad31f554df0bbeff043a7b1b52869b","_id":"@cognitiondesk/mcp@0.5.0","bugs":{"url":"https://github.com/Mounaji-Webs/mcp-cognitiondesk/issues"},"_nodeVersion":"24.18.0","_npmVersion":"11.16.0","dist":{"integrity":"sha512-QSyHJLI1NwzVJD3qd+qAM7/BmBERy3QuKkkzpI8xEfqKdl61qAfGkR5diALF6yMKxOCCiWcMg/16YjgHde9oOg==","shasum":"c592cb127d79e5086494eb6e8d927c9567cda9e5","tarball":"https://registry.npmjs.org/@cognitiondesk/mcp/-/mcp-0.5.0.tgz","fileCount":27,"unpackedSize":160017,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEQCIH2H6N6bOIbBwe0M/99dn5XqNw/QTa2+ALGVAuFNYz4qAiAwAERdh86th687HspFE3PFKmRfIb8KQxWIk7bO0UaVLA=="}]},"_npmUser":{"name":"lucilo","email":"lucilo@gmail.com"},"directories":{},"maintainers":[{"name":"lucilo","email":"lucilo@gmail.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/mcp_0.5.0_1785633347355_0.4753685102223353"},"_hasShrinkwrap":false}},"time":{"created":"2026-08-02T01:15:47.120Z","0.5.0":"2026-08-02T01:15:47.565Z","modified":"2026-08-02T01:15:47.902Z"},"maintainers":[{"name":"lucilo","email":"lucilo@gmail.com"}],"description":"MCP server for the CognitionDesk platform: create and manage AI assistants, workflows, squads and knowledge, and inspect your usage — authenticated with your personal API key.","homepage":"https://app.cognitiondesk.com","keywords":["mcp","modelcontextprotocol","cognitiondesk","ai-assistants","workflows","claude"],"repository":{"type":"git","url":"git+https://github.com/Mounaji-Webs/mcp-cognitiondesk.git"},"bugs":{"url":"https://github.com/Mounaji-Webs/mcp-cognitiondesk/issues"},"license":"MIT","readme":"# @cognitiondesk/mcp\r\n\r\nMCP server for the [CognitionDesk](https://app.cognitiondesk.com) platform. Connect Claude (or any MCP client) to your CognitionDesk account to create and manage AI assistants, workflows, squads and knowledge — and inspect your usage — from your AI tool of choice.\r\n\r\n> **Requires a paid CognitionDesk plan** (Pro or Business) and your own API key. Get it in the app under **Dashboard → API keys**.\r\n\r\n## Quick start\r\n\r\n### Claude Code\r\n\r\n```bash\r\nclaude mcp add cognitiondesk -e COGNITIONDESK_API_KEY=your_key_here -- npx -y @cognitiondesk/mcp\r\n```\r\n\r\n### Claude Desktop\r\n\r\nAdd to your `claude_desktop_config.json`:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"cognitiondesk\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"@cognitiondesk/mcp\"],\r\n      \"env\": {\r\n        \"COGNITIONDESK_API_KEY\": \"your_key_here\"\r\n      }\r\n    }\r\n  }\r\n}\r\n```\r\n\r\nThen ask things like *\"list my assistants\"*, *\"create a customer-support assistant for my store\"*, or *\"how much of my plan have I used this month?\"*.\r\n\r\n## Tools\r\n\r\n| Area | Tools |\r\n|---|---|\r\n| Connection | `get_connection_status` |\r\n| Assistants | `list_assistants`, `get_assistant`, `create_assistant`, `update_assistant`, `delete_assistant`, `try_assistant`, `publish_assistant`, `assign_workflow_to_assistant` |\r\n| Workflows | `list_workflows`, `get_workflow`, `create_workflow`, `update_workflow`, `run_workflow`, `publish_workflow`, `list_workflow_versions`, `get_workflow_analytics` |\r\n| Squads | `create_squad`, `list_squad_templates`, `create_squad_from_template` |\r\n| Knowledge | `search_knowledge`, `list_knowledge_entries`, `get_knowledge_stats`, `ingest_knowledge_text`, `ingest_knowledge_url`, `ingest_knowledge_github`, `delete_knowledge_entry` |\r\n| Skills | `list_skills`, `get_skill`, `resolve_skills`, `create_skill`, `update_skill`, `toggle_skill`, `delete_skill` |\r\n| Custom tools | `list_custom_tools`, `get_custom_tool`, `create_custom_tool`, `update_custom_tool`, `delete_custom_tool` |\r\n| Contacts & identity | `list_assistant_contacts`, `upsert_assistant_person`, `delete_assistant_person`, `upsert_assistant_contact`, `delete_assistant_contact`, `get_identity_audit` |\r\n| Channels | `list_channels`, `list_channel_types` |\r\n| Inventory | `list_inventory`, `import_inventory_items`, `list_import_sources`, `get_import_status`, `sync_import_source` |\r\n| Tasks | `list_tasks`, `get_task`, `create_task`, `update_task`, `delete_task` |\r\n| Usage & plans | `get_usage`, `get_usage_analytics`, `check_action_permission`, `get_available_plans`, `list_scopes` |\r\n| API keys | `list_api_keys`, `create_api_key`, `update_api_key`, `revoke_api_key`, `delete_api_key`, `get_api_key_stats` |\r\n\r\n## Configuration\r\n\r\n| Env var | Default | Description |\r\n|---|---|---|\r\n| `COGNITIONDESK_API_KEY` | — | Your personal API key (**required**; without it the server starts in offline mode and every tool returns a clear error instead of failing silently). |\r\n| `COGNITIONDESK_URL` | `https://mounaji-backendv3.onrender.com` | Backend URL. Only override for local development. |\r\n\r\n## How access actually works\r\n\r\nEverything goes through `/platform-api/*` authenticated with your key, so **your key is the boundary** — not this package:\r\n\r\n- **Scopes.** A key carries scopes (`assistants:*`, `workflows:*`, `wizard:*`, `usage:read` by default). A tool whose scope your key lacks answers `403`, and that's the key's doing, not a bug here. `apikeys:manage` is *not* in the defaults, so a key cannot mint keys unless you explicitly grant it.\r\n- **Assistant scoping.** If your key is restricted to specific assistants (`allowedAssistantIds`), tools only reach those.\r\n- **Plan.** On a free plan the backend answers `403 PLAN_REQUIRED` — upgrade at [cognitiondesk.com/pricing](https://app.cognitiondesk.com/pricing).\r\n- **Metering.** Every call counts toward your plan's quota, and running workflows or ingesting knowledge consumes credits. `get_usage` tells you where you stand.\r\n\r\nUse your own key, never a borrowed one: a key issued by someone else usually reaches *all* of their assistants, including the ones answering real customers.\r\n\r\n## License\r\n\r\nMIT © Mounaji\r\n","readmeFilename":"README.md","_rev":"1-900e981e5c5a78b4e852a83995cea688"}