Workspace
An audit of every keyword's signal against its noise, with exclusions and exact-match changes proposed for your approval.
Copy it into your agent, replace the parts in braces, and save it as a recurring task.
Use the Soundev MCP tools. 1. Call get_usage and list_keywords. 2. For every keyword with health "noisy" (or over 40% low relevance), call search_mentions with keyword_ids [its id], relevance ["low"], include_filtered true, limit 50. 3. Work out what the term is being confused with (another product, a common word, job posts, a person's name). 4. Propose, per keyword: exclude words, must-contain words, or exact match. Estimate how many of those 50 mentions each change would have removed. 5. Return a table and wait for my approval. Do not change any keyword yourself.
Add Soundev as an MCP server in Claude, ChatGPT, Cursor or any MCP client, then sign in when it asks. The address is:
https://soundev.com/api/mcp
Clients without sign-in can use an API key instead (Settings → API keys). Setup details