{
  "markdown": "# Modem MCP Server\n\nConnect [Modem](https://modem.dev) to any MCP-compatible client. Modem builds a custom context graph of your customers and product, grouping disparate data from chat, issue trackers, and support inboxes into topics, people, and companies. From your assistant you can search that graph, run the Modem Agent, and update your workspace directly.\n\n```text\nhttps://mcp.modem.dev/mcp\n```\n\n- **Transport**: Streamable HTTP\n- **Auth**: OAuth 2.1 with Dynamic Client Registration — no API key to create or paste\n- **Docs**: https://modem.dev/docs/api/modem-mcp-server\n\n> This repo holds the public manifests, install docs, and listing assets for the Modem MCP server. The server itself is a hosted service; there is nothing to run from this repo.\n\n## What you can do\n\n| Kind | Tools | Scope |\n| --- | --- | --- |\n| Run the Modem Agent | `modem_agent_invoke`, `modem_agent_get_run`, `modem_agent_send_message`, `modem_agent_cancel_run` | `agent:invoke` |\n| Search your data | `search_modem` — natural-language search over topics, messages, people, companies | `data:read` |\n| Update your workspace | `update_topic`, `bulk_update_topics`, `merge_topics`, `create_companies`, `update_companies`, `merge_companies`, `add_people_to_company`, `update_people`, `merge_people` | `agent:invoke` |\n\nWrite tools act as **you**, under your existing role in the organization you authorize — never more access than your Modem account already has. All writes are annotated destructive, so most clients confirm each call.\n\n## Install\n\nIn every client, the server URL is `https://mcp.modem.dev/mcp`. The first connection opens a browser window for OAuth authorization.\n\n### Claude Code\n\n```bash\nclaude mcp add --transport http modem https://mcp.modem.dev/mcp\n```\n\nThen run `/mcp` inside Claude Code and complete the browser authorization flow.\n\n### Cursor\n\n**Settings** → **Cursor Settings** → **Tools & MCP** → add a new MCP server, or add to `mcp.json`:\n\n```json\n{\n    \"mcpServers\": {\n        \"modem\": {\n            \"url\": \"https://mcp.modem.dev/mcp\"\n        }\n    }\n}\n```\n\nIf Cursor asks for a transport type, choose **Streamable HTTP**.\n\n### VS Code / GitHub Copilot\n\nIn your user profile `mcp.json` or workspace `.vscode/mcp.json`:\n\n```json\n{\n    \"servers\": {\n        \"modem\": {\n            \"type\": \"http\",\n            \"url\": \"https://mcp.modem.dev/mcp\"\n        }\n    }\n}\n```\n\nOr use **MCP: Add Server** from the command palette.\n\n### Codex\n\n```bash\ncodex mcp add modem --url https://mcp.modem.dev/mcp\ncodex mcp login modem\n```\n\nOr in `~/.codex/config.toml`:\n\n```toml\n[mcp_servers.modem]\nurl = \"https://mcp.modem.dev/mcp\"\n```\n\nIf you edit `config.toml` directly, you still need `codex mcp login modem` to complete OAuth.\n\n### Gemini CLI\n\nThis repo is a Gemini CLI extension (see [`gemini-extension.json`](gemini-extension.json)):\n\n```bash\ngemini extensions install https://github.com/modem-dev/mcp\n```\n\nThen authorize when prompted on first use (or run `/mcp auth modem` inside Gemini CLI).\n\n### opencode\n\nIn `opencode.json`:\n\n```json\n{\n    \"$schema\": \"https://opencode.ai/config.json\",\n    \"mcp\": {\n        \"modem\": {\n            \"type\": \"remote\",\n            \"url\": \"https://mcp.modem.dev/mcp\"\n        }\n    }\n}\n```\n\nThen `opencode mcp auth modem` (or authorize when prompted on first use).\n\n### Any other MCP client\n\n| Field | Value |\n| --- | --- |\n| Name | `modem` |\n| URL | `https://mcp.modem.dev/mcp` |\n| Transport | Streamable HTTP (or \"HTTP\") |\n| Authentication | OAuth |\n\n## Example prompts\n\n- \"What are customers saying about billing in the last month?\"\n- \"Summarize the three highest-priority open topics and who's affected.\"\n- \"Mark the topic about SSO login failures as high priority in Modem.\"\n- \"Find the duplicate Acme companies in Modem and merge them into the one with the acme.com domain.\"\n\n## Security and access\n\n- OAuth tokens are scoped to the account and organization selected on the consent screen; the server resolves the organization from token claims, never from the URL.\n- Two permissions, granted separately: `data:read` (natural-language search only) and `agent:invoke` (agent runs + workspace writes).\n- Tool calls are rate limited per organization, per tool (20/min).\n- Privacy policy: https://modem.dev/privacy-policy · Terms: https://modem.dev/terms-of-service · Trust center: https://trust.modem.dev\n\n## Support\n\nQuestions or issues: [support@modem.dev](mailto:support@modem.dev)\n\n---\n\nPlugin packages for Cursor and Claude Code live in [`plugins/`](plugins/).\n",
  "bytes": 4496,
  "sha": "e15e5bb301efa8270ff36f25e15c7250e6ca8d107501c617a824c4df989b192c",
  "repo_slug": "modem-dev/mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_modem_dev_mcp_68aa0967/readme"
}