Skip to content
This repository was archived by the owner on Oct 3, 2026. It is now read-only.
rawtreedbPublic archive

About

RawTree plugin for Cursor: query and ingest data via the hosted RawTree MCP

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

Repository files navigation

This plugin now lives in rawtreedb/agent-skills. This repo is archived.

RawTree for Cursor

Cursor plugin that connects agents to RawTree through RawTree's official hosted Model Context Protocol server, and bundles the official RawTree agent skill.

RawTree is a schema-free OLAP database for agents. Send it raw JSON, events, or telemetry; a table is created on the first insert; read the data back with read-only SQL. This plugin lets Cursor explore your organizations, clusters, databases, and tables, ingest JSON events, run SQL, and read RawTree's query and insert logs from chat.

RawTree account required. The plugin only connects Cursor to RawTree; you need your own RawTree account to sign in. RawTree's website currently offers a Request access flow (rawtree.com/waitlist) and RawTree's launch posts describe a private beta, so access may be limited. If you have no account, the sign-in step will not complete. Contact: contact@rawtree.com.

What's included

Component Path What it is
MCP server mcp.json URL-only connector to https://mcp.rawtree.com/mcp (Streamable HTTP, OAuth), declared with "placement": "server" (hosted MCP path, like other OAuth connectors in cursor/plugins).
Skill skills/rawtree/ The official RawTree skill: how to use the MCP tools, CLI, and API, write read-only SQL, handle Dynamic (schema-free) fields, and tune performance.

Install

  1. Open Cursor Settings → Plugins.
  2. Search for RawTree.
  3. Click Install, then complete the RawTree sign-in prompt.

Or run /add-plugin rawtree in chat.

MCP

{
  "mcpServers": {
    "rawtree": {
      "type": "http",
      "url": "https://mcp.rawtree.com/mcp",
      "placement": "server"
    }
  }
}

Auth is OAuth. Your MCP client opens RawTree in a browser, you sign in and approve access, and no API key or header is needed (RawTree MCP docs). RawTree's OAuth server uses dynamic client registration for public clients and Authorization Code with PKCE (S256); there is no client secret (rawtree-platform docs/auth.md). Each user signs in with their own RawTree account. One connection can reach every organization, cluster, and database the signed-in user can access; resource context is passed per tool call (organization, cluster, optional database) rather than in the MCP URL.

Security: this connection has broad access

The OAuth grant gives the agent broad access to your RawTree account, including destructive operations. RawTree's own docs say: "The OAuth connection grants broad RawTree access, including destructive operations." Tools such as delete-table, delete-database, delete-api-key, create-api-key, remove-organization-member, pause-cluster, and uninstall-app are exposed. The RawTree API enforces your role and resource access on every call, but treat the connection as having your full permissions.

  • Review every write or admin tool call before approving it; confirm the exact organization, cluster, database, and table. Deletions cannot be undone.
  • Be mindful of prompt injection: data you query or ingest and log text returned by list-logs are untrusted input to the agent.
  • Do not paste API keys into chat. create-api-key returns a key that is shown once.
  • You can revoke the connection at any time in RawTree under Profile → OAuth apps (per RawTree's MCP docs), and remove the plugin in Cursor Settings → Plugins. According to RawTree's auth notes, grant revocation takes effect immediately on the API, even for an unexpired access token.

What the OAuth grant covers

RawTree currently publishes a single OAuth scope, full_access (rawtree-platform docs/auth.md). It allows agent-eligible, non-billing operations: organizations, members, clusters, databases, API keys, ingestion, queries, logs, usage, and destructive resource actions. Cluster creation and resizing can incur charges; RawTree's consent screen warns about this. It does not allow billing, payment or subscription operations, organization spending-limit changes, RawTree sign-in/session operations, OAuth grant management, or staff admin routes. Access tokens last 15 minutes; refresh tokens rotate and have a sliding 60-day lifetime.

Tools

The hosted server is the source of truth for tool names and schemas. This list matches the 38 tools registered in rawtreedb/rawtree-mcp at commit 1a37193 (the commit the hosted worker in rawtree-platform pins), and the tools the hosted server listed on 2026-10-02. RawTree's public MCP docs page lists fewer tools, so treat it as a subset.

Category Tools
Discovery list-organizations, list-clusters, get-cluster, list-databases, list-tables, describe-table, list-cluster-sizes
Query run-query (read-only SQL; mutating statements are rejected)
Ingest insert-json (one object or an array; creates the table on first insert), insert-from-url (public JSON/JSONL URL)
Logs list-logs (recent insert/select/describe/explain activity with errors and hints)
Tables and databases create-table, update-table (sorting key), delete-table, create-database, delete-database, verify-database-s3-access
Clusters create-cluster, update-cluster, pause-cluster, resume-cluster, verify-cluster-s3-access
Connectors (Kafka) list-connectors, get-connector, get-connector-metrics, create-connector, add-connector-destination, set-connector-status
API keys list-api-keys, create-api-key, delete-api-key
Apps list-apps, install-app, uninstall-app
Organization members list-organization-members, add-organization-member, update-organization-member, remove-organization-member

Which tools need organization admin

Per the tool descriptions in rawtree-mcp (the API makes the final authorization decision), these state that organization-admin access is required: create-table, update-table, create-database, verify-database-s3-access, verify-cluster-s3-access, create-cluster, update-cluster, pause-cluster, resume-cluster, install-app, uninstall-app, create-connector, add-connector-destination, set-connector-status, add-organization-member, update-organization-member, and remove-organization-member. delete-table is described as requiring an admin key, and RawTree's API rejects it for non-admin credentials.

Read-style tools (for example list-organizations, list-clusters, get-cluster, list-apps, list-organization-members, list-connectors) need a signed-in user with organization membership. The descriptions of the other tools (queries, inserts, logs, delete-database, and the API-key tools) do not state an organization-admin requirement for OAuth users; the API-key tool descriptions refer to admin API-key permission. Whatever the tool says, the RawTree API enforces your role. Destructive and provisioning tools tell the agent to confirm the exact target with you before calling them.

Example prompts

  • "List my RawTree organizations, clusters, and databases."
  • "Show the tables in my default RawTree database and describe the events table."
  • "Insert this JSON event into a RawTree table called app_events, then describe the table."
  • "Query RawTree: daily event counts for the last 7 days from app_events, with a LIMIT."
  • "Check my RawTree logs for failed inserts in the last hour and explain the errors."
  • "Load this public JSONL file into a RawTree table and verify the row count."
  • "Review this RawTree SQL query for correctness and performance."

Notes

  • Features, tools, and limits may change; RawTree's access policy is set by RawTree.
  • Data you ingest and query is stored in your RawTree cluster. Usage terms and billing are governed by RawTree's own terms.
  • For headless agents that cannot open a browser, RawTree's docs describe sending a RawTree API key as a Bearer token to the same hosted endpoint (Authorization: Bearer rt_...). Such calls may omit organization, cluster, and database to use the key's bound cluster, and user-level tools such as list-organizations require OAuth. This plugin uses OAuth only and does not store or configure keys.

Links

Attribution and license

This plugin is authored and maintained by gnzjgo (@gnzjgo, gonzalo@tinybird.co). The bundled skill in skills/rawtree/ is copied from rawtreedb/agent-skills (Copyright 2026 RawTree, Apache-2.0), and the hosted MCP server is operated by RawTree. This plugin is licensed under the Apache License 2.0. The RawTree name and logo are trademarks of their owner; the logo is RawTree's mark from rawtreedb/rawtree-platform (web/public/favicon.svg).

About

RawTree plugin for Cursor: query and ingest data via the hosted RawTree MCP

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors