The Twitter MCP serverwithout the per-tweet API bill
Octolens monitors X for your keywords — tweets and replies — filters the noise with AI, and serves clean mentions to Claude, ChatGPT, and Cursor over MCP. No X API key, no pay-per-read metering, no bot spam in your results.
5-minute setup • OAuth sign-in, read-only • MCP on every planHot take after two weeks of eval: the expensive one is expensive because the filtering actually works. Thread below 🧵
Post · 89 reposts · 214 likes · 4h ago
Reply: “anyone got a recommendation for keyword monitoring that doesn't cost more than the intern who used to do it?”
Reply · under a viral build-in-public post
Trusted by brands people talk about online
X Made Reading Tweets a Metered Service
The conversations are still there — the affordable ways to watch them programmatically are not.
Reads are billed per tweet
X's API now bills reads per post — around half a cent each, with no free read tier. A DIY monitor pulling 50,000 tweets a month is paying for every one of them, including the 90% that turn out to be noise.
Raw results are bot soup
Keyword streams off X arrive unfiltered: engagement bots, spam rings, name collisions, giveaway accounts. Without relevance scoring you pay to read them — then read them again yourself.
The mentions that matter don't tag you
Buying questions and verdicts happen in replies and quote posts where nobody @-mentions your handle. Notifications catch the polite mentions; the revenue-relevant ones you have to go find.
Octolens runs X collection on its side — tweets and replies, continuously — scores every match with AI, and your assistant gets the mentions worth acting on at a flat plan price.
From Viral Thread to Answer in Your AI Tool
One pipeline: Octolens does the collection and filtering, MCP does the delivery. You just ask.
Add keywords
Your brand, handles, competitors, and the category phrases your buyers use — plus negative keywords for noise.
octolens.com → KeywordsOctolens collects
Tweets and replies matching your keywords, with a 7-day backfill per keyword and continuous collection after. Hourly refresh on Pro, real-time on Scale.
tweets + repliesAI scores everything
Every mention arrives scored for relevance and sentiment, tagged with labels like buy_intent, competitor_mention, or complaint — bots filtered out.
"relevanceScore": "high"Your assistant queries it
23 MCP tools: list and search mentions, run analytics, manage keywords and feeds — from Claude, ChatGPT, or Cursor.
list_relevant_mentionsFull tool reference in the MCP docs. The same data is available over the REST API and webhooks — on every plan. Prefer alerts and a feed over an AI assistant? See Twitter monitoring.
Connected in Five Minutes
No X account or API key at any step. You need an Octolens account (7-day free trial) and an MCP-compatible tool.
Add your keywords in Octolens
Brand, handles, competitors, category terms. X collection (tweets + replies) starts with a 7-day backfill.
Connect your client
One URL for every tool. Authentication is a browser sign-in to your Octolens account — there's no API key or token to paste.
Ask
Try: “show twitter mentions from today” — you should see your first scored mentions come back.
claude mcp add --transport http octolens \
"https://app.octolens.com/api/mcp/v2"
# then, inside Claude Code:
/mcp # → octolens → AuthenticateRun the command, then run /mcp inside Claude Code, select octolens → Authenticate, and sign in when the browser opens.
Verify it works
@octolens show twitter mentions from todayNew keywords can take up to 24 hours to index fully — check mentions in the Octolens dashboard first if nothing comes back.
What You'll Actually Ask It
Copy-paste these into Claude, ChatGPT, or Cursor once you're connected.
7-day free trial · No credit card required
MCP Is One Door Into the Same Data
Octolens is the data layer for what people say online. Your assistant queries it over MCP — your team and your agents can consume it everywhere else too.
Webhooks → your stack
Push instead of poll: scored Twitter/X mentions delivered to Slack or your own endpoint as they're found. Wire buying-intent posts straight into your CRM.
Explore webhooks →Agent skills
Ready-made playbooks agents can run over the MCP server — mentions to Linear issues, churn signals to CS, public conversations to pipeline.
Browse agent skills →API → your product
The same Twitter/X data over one REST endpoint with one schema across the internet. Build monitoring into your own product or internal tools.
Build with Octolens →Your Three Options for X via MCP
X's API is pay-per-use now, so every option has a real cost. Here's the honest version.
| Octolens MCP | DIY X-API MCP | Manual scrolling | |
|---|---|---|---|
| X API key required | No | Yes — pay-per-read billing | No |
| Cost model | Flat plan price | ~$0.005 per tweet read, noise included | Free + hours every week |
| Replies & quote-post coverage | Yes | If your queries think to ask | Only what the algorithm shows you |
| Bot & spam filtering | Built-in | You build it — after paying to read the spam | Your eyeballs |
| AI relevance & sentiment scoring | Built-in | You build it | Your eyeballs |
| Historical data & trends | Yes | Only what you've stored | No |
| Other platforms via the same server | All sources | X only | — |
| Cost | $199/mo ($159/mo annual) | Your time + metered reads | Free + your attention span |
A DIY X MCP means paying per read for unfiltered data and building the filtering yourself. Manual scrolling means the algorithm curates your brand intelligence. If X drives real conversations about you, monitor it properly.
Twitter/X MCP Server & Monitoring, Answered
Setup, costs, and coverage — everything about monitoring X from Claude, ChatGPT, or Cursor.
No. Octolens handles X data access on its side — you configure keywords and connect the MCP server with a browser sign-in to your Octolens account. No API application, no pay-per-read billing, no rate limits to manage.
Tweets and replies matching your keywords. Reply coverage matters: buying questions and candid verdicts usually happen in reply threads and quote posts where nobody tags your handle.
A DIY server wraps X's API, which now bills per post read with no free read tier — you pay for the noise, then build filtering, storage, and alerting yourself. Octolens serves pre-filtered, scored mentions from a pipeline that's already running, at a flat plan price.
Same platform. We say 'Twitter' where people still search for it and 'X' where the product does — Octolens monitors x.com either way, and all features work the same.
Depends on your plan: hourly refresh on Pro ($199/mo, or $159/mo billed annually), real-time on Scale ($599/mo, or $499/mo billed annually). If you care about catching viral moments as they start, Scale's real-time refresh is the one.
Yes, in your prompts — ask for mentions from accounts above a follower threshold, or sorted by engagement, and your assistant filters the results. Useful for deciding what deserves a reply this hour versus this week.
Claude Code, Claude Desktop and Web, ChatGPT, Cursor, and any other MCP-compatible client. It's one remote server URL with OAuth sign-in — there's no API key to manage and nothing to run locally.
Every new keyword gets a 7-day backfill, then collects continuously — your archive builds from the day you add a keyword. That's enough history for trends, sentiment shifts, and share-of-voice over time.
Give your AI tools ears on X
Add your keywords, connect the MCP server, and ask your assistant what X is saying about you — all inside the 7-day free trial.
No credit card required · API access on all plans · Cancel anytime