Optimize your Octolens keywords

Set it up

https://app.octolens.com/api/mcp/v2

Run the keyword audit playbook below once. Lookback: {LOOKBACK_DAYS} days. Show me the audit before changing anything.

What it does

• ✓Checks your mention usage against plan limits
• ✓Scores every keyword: volume, high-relevance share, sentiment
• ✓Samples the noise on flagged keywords and proposes exclusions, context, or pausing
• ✓Reviews pending AI keyword suggestions with accept/reject recommendations
• ✓Suggests missing keywords: misspellings, feature names, untracked competitors
• ✓Applies changes only after you approve — and never deletes without being asked

Mentions come from every platform Octolens monitors:RedditX (Twitter)LinkedInHacker NewsGitHubYouTubeBlueskyStack OverflowTikTokDEV.toNewsNewslettersPodcastsWeb

Delivers: A keyword audit with a keep/tune/pause verdict per keyword, proposed fixes with reasoning, and approved changes applied — fewer wasted mentions, better results.

Inputs

How your agent runs it

1. 1

   Map the workspace — Usage vs plan limits, every keyword's config, and the global filters.

   get_usagelist_keywordsget_global_filters
2. 2

   Score signal-to-noise — Per-keyword volume, high-relevance share, and sentiment over the lookback window.

   analytics_contextanalytics
3. 3

   Diagnose the noisy ones — Samples low-relevance mentions to find the off-topic patterns and proposes the fix.

   list_mentions
4. 4

   Review AI suggestions — Accept/reject recommendation for every pending keyword suggestion.

   list_keyword_suggestions
5. 5

   Report and wait — Posts the audit — verdicts, fixes, gaps — and stops. Nothing changes without approval.

6. 6

   Apply approved changes — Executes only what you approved. Deletions require an explicit ask.

   update_keywordpause_keywordadd_keywordaccept_keyword_suggestion

Tools used

• get_usageOctolens MCPCurrent usage against plan limits
• list_keywordsOctolens MCPEvery monitored keyword with platforms, tag, and status
• get_global_filtersOctolens MCPWorkspace-wide filters applied across all queries
• analytics_contextOctolens MCPAnalytics schema and valid filter values
• analyticsOctolens MCPPer-keyword volume, relevance share, and sentiment via SQL
• list_mentionsOctolens MCPSample low-relevance mentions to diagnose noise
• list_keyword_suggestionsOctolens MCPPending AI-proposed keyword tweaks
• update_keywordOctolens MCPApply approved exclusions, context, or platform changes
• pause_keywordOctolens MCPPause approved keywords without deleting them
• add_keywordOctolens MCPAdd approved new keywords

All 23 Octolens MCP tools are documented on the MCP page and in the docs.

The full playbook

Replace the placeholders, then paste this into your agent.

Do the following once (run it monthly, or whenever your feed feels noisy):

1. Get the lay of the land:
   - Call get_usage for plan limits and how much of the mention budget is used.
   - Call list_keywords for every keyword with platforms, tag, and status.
   - Call get_global_filters for workspace-wide filters already in place.

2. Measure each keyword's signal-to-noise. Use analytics_context, then the
   analytics tool to compute, per keyword, over the last {LOOKBACK_DAYS} days:
   - total mentions
   - share of High relevance vs Low/unscored
   - sentiment split
   Flag keywords that eat budget: high volume with under ~30% high relevance.

3. Diagnose each flagged keyword. Call list_mentions and sample ~20 of its
   low-relevance mentions. Look for recurring off-topic patterns: other
   products with the same name, unrelated common-word matches, spammy
   sources. For each, propose a fix: exclusion terms, added context,
   a narrower platform list, or pausing the keyword.

4. Review pending AI suggestions with list_keyword_suggestions. Recommend
   accept or reject for each, with one line of reasoning.

5. Look for gaps worth tracking: common misspellings, product and feature
   names, competitor names not yet covered. Propose additions and note the
   expected volume based on what you saw in the mentions.

6. Post the audit in chat:
   - Usage: mentions used vs plan limit
   - Table: keyword | mentions ({LOOKBACK_DAYS}d) | % high relevance | verdict (keep / tune / pause)
   - Proposed changes, each with its reasoning
   - Suggested additions

7. Only after the user approves specific changes: apply them with
   update_keyword, pause_keyword, add_keyword, accept_keyword_suggestion,
   or reject_keyword_suggestion. Never call delete_keyword unless the user
   explicitly asks for a deletion.