Step 1 — Gather context
If references/learnings.md exists, read it first for accumulated platform knowledge.
Ask the user:
- What do you need help with?
- A) Setting up monitoring (keywords, sources, alerts)
- B) API or webhook integration
- C) MCP server setup for Claude/Cursor/Windsurf
- D) Mention quota management (running out too fast)
- E) AI relevance scoring or sentiment issues
- F) Multi-project or team configuration
- G) Something else — describe it
- Which plan are you on?
- A) Free trial (Pro features, 1K mentions)
- B) Pro ($119/mo — 15K mentions, 10 keywords)
- C) Scale ($319/mo — 50K mentions, 15 keywords)
- D) Enterprise (custom)
- E) Not sure / evaluating
If the user's request already provides enough context, skip directly to Step 2.
Step 2 — Route or answer directly
If the request maps to a strategy skill, route:
- Social listening tool comparison or strategy →
/sales-social-listening [question] - Reddit-specific deep monitoring (subreddit analysis, SERP tracking) →
/sales-threadlytics [question] - AI visibility monitoring (LLM brand mentions) →
/sales-ai-visibility [question] - Buyer intent signals from social mentions →
/sales-intent [question]
Otherwise, answer directly from the platform reference below.
Step 3 — Octolens platform reference
Read references/platform-guide.md for the full platform reference — capabilities, pricing, data model, API details, MCP setup, integration recipes, code examples.
Answer the user's question using only the relevant section. Don't dump the full reference.
Step 4 — Actionable guidance
Focus on the user's specific situation:
- Keyword optimization — maximize coverage within limits using variations and Boolean-like strategies
- Mention budget — configure filtered views, use AI scoring to reduce noise
- Integration — API, webhooks, MCP, or Zapier/Make/n8n setup
- Team workflow — filtered views, Slack routing, escalation rules
If you discover a gotcha or tip not covered in references/learnings.md, append it there.
Gotchas
Best-effort from research — review these, especially items about plan-gated features and integration gotchas that may be outdated.
- 10-15 keyword limit is hard. Pro gives 10, Scale gives 15. Use compound keywords and variations within a single keyword slot rather than burning one per variation. Extra keywords cost $5-10/mo each.
- Mentions are consumed on match, not on view. Even irrelevant matches count against your quota. Tight keyword context and filtered views reduce waste but don't prevent consumption.
- Hourly vs real-time refresh. Pro plan refreshes hourly. Scale refreshes in real-time. If you need instant alerts for crisis detection, you need Scale.
- No LinkedIn engagement intelligence. Octolens tracks LinkedIn mentions (keyword matches in posts) but doesn't provide engagement analytics, profile changes, or competitor content interaction data.
- MCP key is different from API key. Generate MCP keys at Settings > MCP, not Settings > API. Using the wrong key type causes silent auth failures.
- Links in content trigger false positives. If your keyword appears inside a URL in a post, it may match even though the post isn't about your brand.
Related skills
/sales-social-listening— Social listening strategy — brand monitoring, sentiment analysis, competitive intelligence, crisis detection, tool comparison/sales-threadlytics— Threadlytics platform help — Reddit-specific monitoring with 500M+ conversations, opportunity scoring, SERP tracking/sales-brand24— Brand24 platform help — affordable social listening with MCP server, Storm Alerts/sales-awario— Awario platform help — budget social listening with Boolean search, Awario Leads/sales-ai-visibility— AI visibility monitoring — track what LLMs say about your brand/sales-do— Not sure which skill to use? The router matches any sales objective to the right skill. Install:npx skills add sales-skills/sales --skill sales-do
Examples
Example 1: Optimize keyword strategy within limits
User says: "I'm on the Pro plan with 10 keywords. I want to track my brand, 3 competitors, and industry terms but that's already 5+ keywords." Skill does:
- Reads platform guide for keyword optimization patterns
- Suggests compound keyword strategies (brand + variations in one slot)
- Recommends using filtered views to segment without extra keywords
- Notes the $5/mo add-on keyword option if optimization isn't enough Result: Maximized monitoring coverage within the 10-keyword limit
Example 2: Set up MCP server for Claude Code
User says: "How do I connect Octolens to Claude Code so I can query mentions from my terminal?" Skill does:
- Reads platform guide MCP setup section
- Provides the
claude mcp addcommand with SSE transport - Explains team key vs personal key choice
- Suggests test query to verify connection Result: Working MCP connection with example queries
Example 3: Build a webhook pipeline to Slack
User says: "I want high-relevance mentions pushed to a Slack channel in real-time instead of checking the dashboard" Skill does:
- Reads platform guide webhook and integration patterns
- Outlines webhook endpoint setup for mention push
- Provides code for filtering by AI relevance tag before forwarding to Slack
- Notes hourly vs real-time refresh plan differences Result: Webhook pipeline filtering high-signal mentions to Slack
Troubleshooting
Mentions running out before month end
Symptom: 70%+ of mention quota consumed in first week Cause: Keywords too broad, matching common terms in irrelevant contexts Solution: Review top mentions — identify recurring false positives. Add context keywords in Octolens to narrow matches. Use filtered views to focus on high-relevance AI tags only. Consider removing low-value keywords and replacing with more specific ones.
MCP server not responding in Claude Code
Symptom: @octolens list my keywords returns no results or connection error
Cause: Wrong key type, expired key, or firewall blocking app.octolens.com
Solution: Verify you're using an MCP key (Settings > MCP), not an API key. Check key expiration date. Ensure outbound connections to app.octolens.com are allowed. Restart Claude Code completely after adding the MCP config.
Webhook not delivering mentions
Symptom: Configured webhook endpoint but no POST requests arriving Cause: Endpoint not publicly reachable, wrong URL, or plan refresh interval Solution: Test endpoint reachability from outside your network. Verify URL in Octolens webhook settings (must be HTTPS). Note that Pro plan refreshes hourly — mentions won't push until the next refresh cycle. Check Octolens webhook delivery logs if available.