Skip to content
v1.0.0MIT

Run AI perception studies from your agent: see how AI models describe a brand, topic or issue, segment by segment, and read share of voice, sources and narratives.

LLMEKNOW plugins

Official LLMEKNOW plugins for Claude, ChatGPT and Codex, and Cursor.

LLMEKNOW measures how AI models describe a brand, a topic or a public issue. This plugin connects your agent to your LLMEKNOW organisation so it can set up a study, run a wave, and read the results: share of voice, top of mind and sentiment per entity, the web sources the AI answers cite, and the narratives they repeat, overall and for each audience segment against a baseline with no segment applied.

The plugin has two parts:

  • A connection to the hosted LLMEKNOW MCP server at https://llmeknow.com/api/mcp. It uses OAuth: the first time a tool is used, your browser opens LLMEKNOW, an owner or admin of your organisation signs in, picks the organisation, can set a 30-day spend cap, and approves. No key goes in any config file.
  • Three skills that teach the agent the workflow:
    • getting-started: checks the account first, explains the vocabulary, and hands human-only steps back to you.
    • run-a-study: brief, market, segments, campaign, cost, your agreement, wave, progress.
    • read-the-results: entities, segments against the baseline, sources, narratives, wave comparisons and raw responses.

You need an LLMEKNOW organisation with agent access switched on. See Connect your own agent.

Install

Claude Code

/plugin marketplace add murmur-intelligence/llmeknow-plugins
/plugin install llmeknow@llmeknow

Or from your shell:

claude plugin marketplace add murmur-intelligence/llmeknow-plugins
claude plugin install llmeknow@llmeknow

Claude (web and desktop)

Settings, Connectors, Add custom connector, with the URL https://llmeknow.com/api/mcp. No client id or secret is needed.

Codex and ChatGPT

codex plugin marketplace add murmur-intelligence/llmeknow-plugins

Then install LLMEKNOW from that marketplace in the Plugins Directory. To connect the server alone, add a connector in ChatGPT with the URL https://llmeknow.com/api/mcp.

Cursor

Clone the repository into Cursor's local plugin folder, then run Developer: Reload Window:

git clone https://github.com/murmur-intelligence/llmeknow-plugins ~/.cursor/plugins/local/llmeknow

To connect the server alone, add {"url": "https://llmeknow.com/api/mcp"} as an MCP server in Cursor's settings.

Scripts and headless agents

An owner or admin creates an API key in LLMEKNOW under Settings, API keys. Send it to the same URL as Authorization: Bearer llmk_.... Keep the key in your own secret store, never in this repository.

Spending

Reading data is free. Running a wave, retrying or resuming one, generating audience members, and backfilling a narrative taxonomy spend from your organisation's prepaid wallet. The skills tell the agent to show you the cost and wait for your explicit yes before any of these. The server enforces the same rule: a wave starts only when the call echoes the current cost estimate, and a connection with a spend cap refuses anything over it.

Sign-up, payment, adding wallet funds, plan changes, campaign slots and API keys stay in the LLMEKNOW web app. When one of these blocks a task, the agent tells you what to do and links to the right page.

Data and privacy

  • This repository contains no executable code that runs in your agent: only manifests, instructions for the agent, and icons. The scripts/ folder holds the repository's own checks and is not part of the plugin's behaviour.
  • The plugin connects to one server, https://llmeknow.com/api/mcp, over HTTPS. It sends nothing anywhere else.
  • LLMEKNOW receives the tool calls your agent makes and their arguments, never your conversation. Each call is recorded against the connection for audit.
  • The connection acts for one organisation, with the permissions of the member who approved it. Revoke it at any time in Settings, API keys; a wave it already started keeps running.
  • Results contain text written by AI models and synthetic audience members. The skills tell the agent to treat that text as evidence, never as instructions.

Privacy policy and terms.

Support

Email support@llmeknow.com or use llmeknow.com/contact. Report problems with the plugin itself as an issue on this repository.

Repository layout

PathFor
.claude-plugin/plugin.json, .claude-plugin/marketplace.json, .mcp.jsonClaude Code and the Claude plugin directory
plugin.json, mcp.jsonAgent Plugins 1.0.0 format, read by ChatGPT, Codex and Cursor; OpenAI listing details under extensions.com.openai
.cursor-plugin/plugin.jsonCursor plugin format
server.jsonMCP Registry entry
skills/The three skills, shared by every platform
assets/Icons
TEST_CASES.mdReview test cases

Licence

MIT. See LICENSE.