Skip to directory

Grep Public API v2 MCP

Grep AI (Parcha, Inc.)

Streamable HTTP26+ documented tools2

About Grep Public API v2 MCP

A cloud-hosted MCP server for starting and managing Grep research runs, reading generated files, managing attachments and agents, and checking quota and billing.

What you can use it for

  • Run sourced research jobs and inspect reports
  • Read artifacts generated by research runs
  • Manage custom agents, skills, and attachments
  • Check quota and billing usage before research
Connection and setup
Hosting
remote
Transport
Streamable HTTP
Endpoint
https://api.grep.ai/mcp
Authentication
Browser-based OAuth 2.0. MCP clients discover the protected resource and authorization server from OAuth metadata. Claude Code and Codex use their documented mcp login flow; users complete sign-in in the browser. API keys are supported by the separate REST API, not substituted into this MCP recipe.

Requirements

  • A Grep account; research runs consume account quota/credits.
  • MCP client supporting remote Streamable HTTP and OAuth browser login.
  • For ChatGPT, workspace administrator must enable Developer mode.

Setup steps

  1. Add remote endpoint https://api.grep.ai/api/v2/mcp using Streamable HTTP.
  2. Start OAuth from the client. Claude Code: claude mcp add --transport http grep https://api.grep.ai/api/v2/mcp, then claude mcp login grep. Codex: codex mcp add grep --url https://api.grep.ai/api/v2/mcp, then codex mcp login grep.
  3. Complete browser sign-in and authorize the requested scopes.
  4. Check quota before starting paid research. Verify with a harmless quota_get call or list visible runs.
Tools 26+

Official MCP server card/docs expose a large evolving v2 inventory across research runs, files, attachments, agents/experts, quota and billing. The stored 26 names are a partial subset, not complete tools/list; OAuth scopes determine access. Check quota/billing before launching potentially chargeable research.

  • run_create

    Start a Grep research run.

    01
  • run_continue

    Continue a completed research run.

    02
  • run_cancel

    Cancel an in-progress run.

    03
  • run_list

    List research runs visible to the OAuth subject.

    04
  • run_get

    Get a run report and state.

    05
  • run_timeline

    List timeline events for a run.

    06
  • run_files_list

    List files generated by a run.

    07
  • run_file_read

    Read a generated research file.

    08
  • attachment_create

    Upload one or more attachments.

    09
  • attachment_get

    Inspect an attachment.

    10
  • attachment_delete

    Delete an attachment.

    11
  • quota_get

    Read current public API quota.

    12
  • quota_check

    Check quota for a requested effort.

    13
  • billing_usage_get

    Read billing usage.

    14
  • agent_list

    List custom agents.

    15
  • agent_create

    Create a custom agent.

    16
  • agent_update

    Update an agent configuration.

    17
  • agent_delete

    Archive a custom agent.

    18
  • skills_list

    List or search skills.

    19
  • skill_get

    Read skill content.

    20
  • skill_create

    Create a custom skill.

    21
  • mcp_tools_list

    List tools available for agent grants.

    22
  • batch_create

    Create a batch of research runs.

    23
  • batch_get

    Read batch status and progress.

    24
  • campaign_create

    Create a campaign draft.

    25
  • campaign_approve

    Approve and start paid campaign work.

    26

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

Limitations and notes
  • Research and campaign actions can consume paid credits; check quota and cost before launching.
  • Some tools mutate state, including creating/canceling runs, attachments, agents, and campaigns; campaign_approve explicitly starts paid work.
  • Tool names evolve; the official server card is the authoritative live catalog.
  • The similarly named grep.app search wrapper is a separate implementation and is not this product.
Original sources
  • official primary

    Grep MCP docs describe public v2 remote Streamable HTTP MCP, browser OAuth, client setup for Claude Code, ChatGPT and Codex, use-case/tool families, and quota/billing cautions.

  • official primary

    Grep authentication docs distinguish OAuth tokens for agents from API keys used for direct API requests and warn to keep keys in environment/secrets storage.

  • official live server card

    Current official MCP server card provides canonical public endpoint, Streamable HTTP and OAuth metadata; the corresponding current catalog is larger and evolving, so stored tool names remain explicitly partial.

Documentation checked 2026-10-28. Connection has not been runtime tested.

Similar tools