contextcore-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@contextcore-mcpLoad the full context for my WebApp project"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ContextCore for AI editors
Official ContextCore plugin for Cursor and Claude Code — product context over MCP, the ContextCore playbook skill, and the init command. Claude Desktop uses Custom Connectors or an optional .mcpb Extension.
Editor matrix: EDITORS.md — Cursor, Claude Code plugin, Claude Desktop Connectors,
.mcpb, stdio, CLI.
Give agents secure access to your ContextCore project context — vision, audience, scope, tech decisions, rules, skills, risks, and open questions — plus write tools to propose updates.
With a read-scoped token or OAuth grant, the agent reads published context only. With write access, it can also propose context, rules, skills, risks, and problems (landing in New for review unless you have direct-publish permission).
The agent gets exactly the context you can see. Access is enforced server-side by ContextCore's org/team/visibility federation.
Quick start
Editor | Action |
Cursor | Customize → search ContextCore → Install |
Claude Code |
|
Claude Desktop | Customize → Connectors → |
Details |
Related MCP server: CollabMCP
Cursor + Claude Code plugins (this repo)
This repo is the ContextCore editor plugin for Cursor and Claude Code — shared skills, commands, and MCP wiring. Claude Desktop uses Custom Connectors (same gateway URL) or the optional .mcpb in desktop-extension/. See EDITORS.md.
Cursor Directory: cursor.directory/plugins/contextcore (live).
Official marketplace: submitted 2026-09-03 — awaiting review.
Claude Code directory: submitted 2026-09-08 — pending review. See SUBMISSION.md.
Local Cursor test: clone to ~/.cursor/plugins/local/contextcore and reload Cursor.
Claude Code test:
claude plugin marketplace add babaVC/contextcore-mcp
claude plugin install contextcore@contextcore-mcpClaude Code (plugin or MCP-only)
Plugin (recommended): see above — bundled skill + /contextcore:contextcore-init.
MCP only:
claude mcp add --transport http contextcore https://cloud.contextcore.md/mcpBrowser OAuth on the first MCP call. Paste Set up ContextCore and initialize it. — same Initialize procedure as /contextcore-init.
More: EDITORS.md · MCP user guide
MCP servers in the plugin (Cursor)
Primary — contextcore (remote / OAuth). No token to paste. The first request opens browser sign-in and consent. Configure CONTEXTCORE_MCP_URL (default https://cloud.contextcore.md/mcp).
Fallback — contextcore-stdio. For headless setups or when you prefer a Personal Access Token. Set CONTEXTCORE_TOKEN and optionally CONTEXTCORE_API_URL. Uses this repo's npm package via npx.
If you only use OAuth, leave CONTEXTCORE_TOKEN unset and disable the contextcore-stdio server in Cursor MCP settings — otherwise it shows as failed until a token is configured.
On-prem: point CONTEXTCORE_MCP_URL and CONTEXTCORE_API_URL at your instance.
What your agent gets
Read tools (default):
Tool | What it returns |
| Projects you can access (id, name, tagline) |
| Resolved map, gaps, counts, watch roll-ups |
| A project's full context as one Markdown document |
| File layout of the Core export |
| One component (e.g. |
| One rule or skill by slug — use |
| Component list for a project |
| Keyword matches across pieces, rules, skills, risks, and problems |
Write tools (write-scoped token or OAuth grant with propose changes):
Tool | What it does |
| Propose a context piece or update (with |
| Propose agent guidance assets |
| Propose watch items |
| List pending items in New |
| Approve or reject a suggestion |
Tools are fetched from the server at startup — no package upgrade needed when the gateway adds tools.
Standalone stdio package (non-Cursor editors)
The contextcore-mcp npm package (bin/cli.js in this repo) is a JSON-RPC proxy for stdio-only MCP clients. The Cursor plugin reuses it as contextcore-stdio; other editors can use it directly:
{
"mcpServers": {
"contextcore": {
"command": "npx",
"args": ["-y", "contextcore-mcp"],
"env": { "CONTEXTCORE_TOKEN": "YOUR_TOKEN" }
}
}
}For sync/review/propose from the terminal, see contextcore-cli.
Manual MCP config (no plugin)
Remote OAuth — works in Cursor, Claude Code, and other HTTP MCP clients:
{
"mcpServers": {
"contextcore": {
"url": "https://cloud.contextcore.md/mcp"
}
}
}Mint a token under Account → Integrations only if you need stdio or automation.
Sync workflow (maintainers)
Playbook copy lives in the main ContextCore app repo, not here. skills/ and commands/ are generated — do not hand-edit.
Edit
src/constants/mcpPlaybook.jsin context-os (private app repo)Run
npm run sync:cursor-pluginfrom that repo
That writes skills/contextcore/SKILL.md, commands/contextcore-init.md, and dogfoods the command into context-os/.cursor/commands/.
Configuration
Variable / env | Required | Default | Description |
| no |
| Remote MCP gateway (plugin variable) |
| no |
| App origin for stdio proxy |
| stdio only | — | Personal Access Token from ContextCore |
Legacy CONTEXTOS_TOKEN / CONTEXTOS_API_URL still work in the stdio client.
Security
OAuth grant or token maps to one ContextCore user. Federation gates match the web app.
Write proposals land in New unless you have direct-publish permission.
Token hashes only — revoke anytime from Account → Integrations.
Never commit tokens. Use plugin variables or
${env:…}in config.
See contextcore.md/trust.
License
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Share one project context across ChatGPT, Claude, Telegram and any MCP client.
Shared, permission-aware company context for AI agents, with provenance, approvals and audit.
- OneLoreOAuthai.onelore
Shared project context for AI agents and teams: docs, tasks, and messages that stay current.
Securely search and manage workspace context files for AI agents and teams.
Related MCP Servers
- AlicenseAqualityDmaintenanceEfficiently loads, caches, and delivers project file context for AI agents using the Model Context Protocol, reducing multiple file reads to a single call with LRU caching and security validation.315 npmMIT
- AlicenseAqualityDmaintenanceProvides shared real-time context for Claude Code agents, including scope awareness, prior decisions, and anti-overlap, using only files in the repo.5MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI tools to retrieve codebase context, summaries, decision notes, and file details via MCP, avoiding re-explaining the project.2MIT
- FlicenseNot gradedqualityBmaintenanceProvides a read-only context gateway for the Mindway operating protocol, enabling AI clients to load startup context, search public guidelines, and retrieve protocol documentation safely with path traversal protection.-