| FMRS | 57 / 100 · C | 66 / 100 · C | 64 / 100 · C |
|---|
| Reliability | 9 / 20 | 8 / 20 | 8 / 20 |
|---|
| Security and permissions | 10 / 20 | 15 / 20 | 17 / 20 |
|---|
| Maintenance | 13 / 20 | 15 / 20 | 10 / 20 |
|---|
| Documentation | 14 / 20 | 15 / 20 | 16 / 20 |
|---|
| Setup experience | 11 / 20 | 13 / 20 | 13 / 20 |
|---|
| Best for | - Teams that want to self-host and fully own their meeting data and transcripts
- Regulated organizations needing air-gapped meeting intelligence on their own infrastructure
- Developers wiring meeting transcripts into their own AI agents and knowledge bases
| - Developers, content creators, and project teams using Atlassian Cloud
- Teams that want Atlassian access inside ChatGPT, Claude, Cursor, or VS Code
- Users who need natural-language search, write, and cross-product workflows
| - Teams and individuals who already manage docs/wikis in Notion and want AI clients (Claude Desktop, Cursor, etc.) to operate on that content directly
- Developers who want to self-host a local MCP server rather than use a remote hosted option
- Automation/agent workflows that need to query or update Notion database records programmatically
|
|---|
| Not for | - Users who only want hosted SaaS and do not need self-hosting (agent-plane features are self-hosted only)
- Users needing full mid-call bot control or TTS speech into calls (those endpoints return 404 in the open-core stack)
- Scenarios requiring guaranteed speaker attribution (heavy crosstalk yields empty speakers on roughly 4–7% of rows)
| - Users without an Atlassian Cloud site or relevant product permissions
- Deployments requiring a self-hosted MCP server
- Users unwilling to grant an AI client access to Atlassian data or actions
- Jira Service Management or Bitbucket Cloud workflows requiring OAuth authentication
| - Users who want the standard-OAuth, zero-JSON-config install with active official support — Notion recommends its newer remote Notion MCP server for that
- Users uncomfortable handing a Notion integration token to a locally-run LLM client
- Use cases needing capabilities the integration deliberately excludes, such as deleting databases
- Long-term production deployments that need guaranteed maintenance, given this repo may be sunset
|
|---|
| Required permissions | - A Vexa API key is required for the remote MCP endpoint (Authorization: Bearer header, secret)
- HTTP API calls use the X-API-Key header
- Self-hosting requires Docker (engine ≥ v26 recommended) or a Kubernetes/OpenShift cluster
- Transcription requires an STT service or hosted token; recordings require object storage such as MinIO
| - Jira and Confluence provide read, write, and search permission groups
- Jira Service Management provides read and write access through API tokens only
- Bitbucket Cloud provides read and write access through scoped API tokens only
- Compass is available through OAuth 2.1 only
- The first OAuth installer for a site must have access to the Atlassian apps requested by the MCP scopes
- Organization administrators can manage domain controls, API-token authentication, app access, and audit logs
| - A Notion internal integration token (NOTION_TOKEN or the Authorization value in OPENAPI_MCP_HEADERS), scoped by that integration's Capabilities (can be limited to read-only)
- Access limited to the specific pages/databases explicitly connected to the integration, not the whole workspace
- A separate bearer auth token when using the Streamable HTTP transport (--auth-token / AUTH_TOKEN)
- Network access to the Notion API
|
|---|
| Risks and side effects | - A leaked API key lets someone send bots, read transcripts and dispatch agents as you
- The bot joins calls as a visible participant, so participant knowledge and consent matter
- Recordings and transcripts may contain sensitive content; self-hosting with your own storage controls the data boundary
- Agents handling untrusted input (email, web pages) must run propose-only with human approval, otherwise writes can be irreversible
- In the open-core stack, mid-call bot config and speak endpoints currently return 404 and should not be relied on
| - AI clients can perform Jira, Confluence, Jira Service Management, Bitbucket, and Compass actions on the user’s behalf
- Prompt injection, indirect prompt injection, and tool-poisoning attacks may cause data exfiltration or unintended changes
- Use least privilege, scoped tokens, and minimal project or workspace access
- Require human confirmation for high-impact or destructive actions and monitor audit logs
| - If the integration is granted broad read/write capabilities and connected to many pages, that content is effectively exposed to the LLM, risking data leakage or unintended edits
- Running with --unsafe-disable-auth removes bearer authentication on the HTTP transport, potentially making it reachable via DNS rebinding from pages a user visits — the README says to only use it on an isolated network
- Multi-tenant token passthrough requires careful TLS/gateway configuration; misconfiguration risks leaking or misrouting Notion tokens between tenants
- The README states this local server may be sunset in favor of the remote Notion MCP, and that issues/PRs here are not actively monitored, so long-term maintenance and security fixes are not guaranteed
- Storing NOTION_TOKEN or OPENAPI_MCP_HEADERS directly in shared client config files risks leaking the secret
|
|---|
| Supported clients | Claude Code, Codex | OpenAI ChatGPT, Claude, Cursor, Visual Studio Code (GitHub Copilot), GitHub Copilot CLI, Google Gemini CLI, Amazon Quick Suite, Local MCP-compatible clients via mcp-remote | Claude Desktop, Cursor, Zed, GitHub Copilot CLI |
|---|
| Tools | 0 | 0 | 9 |
|---|