Give your agent eyeson the entire internet
One tool call returns AI-scored mentions from Reddit, X, LinkedIn, Hacker News, news, podcasts & 10+ more. No platform APIs, no scraping, no auth juggling — one key, structured JSON.
50 free searches • No credit card • Your key never expires# 1. Install the open-source CLI
$ npm install -g octolens
# 2. Sign up — opens the browser, mints your API key
$ octolens signup
# 3. Give your agent its first pair of eyes
$ octolens search "your brand" --days 7 --json
Trusted by brands people talk about online
"Wild AI-native onboarding experience. I literally entered zero input other than vercel.com to get my Octolens feed up and giving me mentions."
Agents Can't Get Social Data Themselves
The data your agent needs lives behind a dozen expensive, auth-walled, rate-limited platform APIs. Octolens replaces all of them with one call.
Building the plumbing yourself
- Twitter/X API alone starts at $200/month — before Reddit, LinkedIn, or news
- Every platform has its own auth flow, rate limits, and breaking changes
- Scrapers break the moment a layout changes
- Raw feeds are 80% noise — your agent burns tokens reasoning over garbage
- A dozen response schemas to normalize before your agent can use anything
With Octolens
- One API key, one schema, 15+ platforms
- Relevance and sentiment pre-scored — your agent reasons over signal, not spam
- Pull mode: on-demand search, results in seconds, mid-task
- Push mode: webhooks wake your agent the moment something happens
- CLI, MCP server, and REST — pick whatever fits your agent's stack
Wire It Into Your Agent in Under Two Minutes
CLI for shell-native agents, MCP for agent frameworks, REST for custom code, a skill for coding agents. Same data, same key, every surface.
npm install -g octolens
octolens signup # opens the browser, mints your API key
octolens search "your brand" --days 7 --jsonOpen source, on npm. Signup opens the browser; on a headless box (a Hetzner server, CI) --no-browser prints a sign-in link you can open from any device — or export OCTOLENS_API_KEY. --json output, documented exit codes, no prompts in non-TTY environments.
Runs Inside OpenClaw, Claude Code & Any Agent Harness
Install the skill once and your agent knows how to use Octolens — including on a schedule, with no human kicking it off.
OpenClaw heartbeat
Add a cron heartbeat and your OpenClaw agent checks your mentions every morning, triages them, and messages you only what needs a human.
# install the skill
openclaw skills install @garrrikkotua/octolens
# openclaw.json
{
"heartbeat": [{
"cron": "0 9 * * 1-5",
"prompt": "Search Octolens for mentions from the last 24h. Draft replies for questions, flag negative threads, skip the noise. Message me anything that needs a human."
}]
}
Claude Code
One skill install and Claude Code picks the right command itself — search, filter, and output land straight in your working session.
npx add-skill octolens/skill
> Find this week's mentions of acme on Reddit and HN,
and draft replies for the top three questions.
⏺ octolens search "acme" --days 7 --json
⏺ 8 relevant mentions · 3 questions found
⏺ Drafts written to replies.md
Model-agnostic by design — works with Claude, GPT, Gemini, Llama, and anything that can run a CLI command or call an MCP tool.
Search Answers 'What Happened.' Monitoring Answers 'Tell Me the Moment It Happens.'
Agents need both — a question they can ask mid-task, and a signal that wakes them up. Octolens is the only social data source built for the pair.
On-Demand Search
Your agent asks a question mid-task — "find mentions of this company from the last 7 days" — and gets AI-scored results back in seconds. No keyword setup, no state, no waiting.
- One CLI command, MCP tool call, or POST
- Results scored for relevance and sentiment
- maxResults caps every call, so cost is predictable
- Empty searches cost nothing
Monitored Keywords + Webhooks
Persistent monitoring watches every platform in real time and wakes your agent the moment something happens — a negative thread, a competitor launch, a support issue.
- Real-time delivery to your endpoint
- Filter by platform, sentiment, or relevance before it fires
- octolens feeds watch streams the same feed in the terminal
- The trigger layer for event-driven agent workflows
Ready-Made Skills Your Agent Can Run
Copy-paste playbooks powered by the Octolens MCP — your agent files issues, finds leads, drafts copy, and ships reports from real conversations. No glue code.
Why Agents Love Octolens
Not a dashboard with an API bolted on. Every surface is designed to be driven by something that isn't human.
Headless by contract
The CLI never prompts, spins, or colorizes in non-TTY runs. Documented exit codes for every failure family, structured error envelopes on stderr, and lists degrade to tab-separated rows — so cut and wc just work.
Structured JSON on every surface
CLI --json, MCP, and REST all return the same schema. No HTML parsing, no scraping heuristics — data your agent can reason over directly.
A skill your agent reads itself
npx add-skill octolens/skill ships a SKILL.md with tools, filters, and best practices. Your agent learns the API without you writing a prompt.
Hosted MCP — nothing to run
One URL with your API key. No local server, no Docker, no process to babysit. Works in Claude, Cursor, Windsurf, and OpenClaw out of the box.
Predictable cost per call
maxResults caps every search, so your agent's worst-case spend is knowable before it runs. Empty searches are free.
Fails loudly, not silently
Out of mentions? A documented 402 your agent can catch and handle — not a mystery timeout at 2am. Spend caps keep runaway loops off your bill.
Search Included on Every Plan
You only pay for mentions — however your agent gets them. A search returns up to 10 mentions, counted against your plan like any other mention. No credits, no separate meter, no surprise line items.
Free developer sandbox
- Sign up from the terminal — octolens signup mints your key
- 50 free searches — one-time, they never expire
- CLI, MCP server, REST API & webhooks
- No credit card required
Pro
Billed annually. Usage beyond your quota is billed at $0.013 per mention — set a spend cap or turn it off in settings.
- 15,000 mentions/month — search and monitoring draw from the same allowance
- Real-time monitored keywords with alerts
- Slack, email & webhook notifications
- Full analytics and 2-year retention
Heavy agent workloads? Compare all plans — Scale includes 50,000 mentions and real-time refresh.
Every Tool Your Agent Can Call
The same capabilities on every surface — pick per workflow, not per vendor.
| Surface | Tool / command | What it does |
|---|---|---|
| CLI | octolens search | Search from the shell — --json for pure structured output |
| CLI | octolens keywords add | Set up persistent monitoring for a brand or topic |
| CLI | octolens feeds watchPaid plans | Live-stream new monitored mentions in the terminal |
| CLI | octolens mentions export | Export filtered mentions to CSV/JSON — a file or straight to stdout |
| CLI | octolens analytics | Volume trends, sentiment splits, and top keywords from the shell |
| MCP | search_mentions | On-demand search across all platforms — scored results in seconds |
| MCP | get_search | Retrieve the results of a previous search by ID |
| MCP | list_mentions | Query your monitored mentions with filters and pagination |
| MCP | analytics | Run SQL over your mention data — trends, sentiment, share of voice |
| REST | POST /api/v2/search | The same on-demand search over plain HTTPS |
| REST | POST /api/v2/mentions | Query monitored mentions — filters, cursor pagination |
| Webhooks | on new mention | Push delivery to your endpoint the moment a mention appears |
Real results from teams using Octolens
"Built on top of the Octolens API, I use Railway as a service to pipe all of the mentions to different channels. Then there's a cron job that does a daily digest for us."
"Wild AI-native onboarding experience. I literally entered zero input other than vercel.com to get my Octolens feed up and giving me mentions."
"A big part of our job is paying attention to what people are saying online - answering questions, correcting misunderstandings, and joining the conversation in real time."
"It's not about replying to everything. It's about showing up in the right conversations and doing it genuinely. Octolens helps us do that at scale."
"Developers tune out advertising, but they do talk online - and that's where our reputation is built."
Agent Questions Answered
Search is included on every plan — results simply count against your monthly mention allowance. A search returns up to 10 mentions by default (configurable with maxResults), so one full search uses about 10 mentions, roughly $0.013 each beyond your plan's quota. Empty searches cost nothing: you pay for data received, not queries.
No. The free sandbox bundle is one-time but has no clock — use it today or in six months. Your API key stays active either way, so an integration you build now still works when you're ready to go to production.
Yes — the Octolens CLI and MCP server are open source on GitHub and published to npm. The CLI is fully headless: --json output on every command, documented exit codes, no prompts in non-TTY environments. Built for agents first, humans second.
Yes. A human signs up once — octolens signup opens the browser and mints a scoped API key — and from then on headless environments authenticate with the OCTOLENS_API_KEY environment variable. No interactive prompts, no config files, so the same commands work in CI, cron jobs, and agent harnesses.
By default it doesn't: paid plans include flex mentions, so data keeps flowing past your quota and additional mentions are billed at your plan's per-mention rate. Worried about a runaway loop? Set a monthly spend cap in settings — collection stops at the cap. If you disable flex mentions entirely, the API returns a documented 402 response your agent can handle gracefully — no silent failures, no mystery errors.
Anything that speaks MCP: Claude, Claude Code, Cursor, Windsurf, OpenClaw, and custom agents. The MCP server is hosted — one URL with your API key, nothing to run. If your stack doesn't speak MCP, use the REST API or the CLI.
Search is pull: your agent asks a question mid-task and gets scored mentions from the recent past in seconds — no setup. Monitored keywords are push: persistent, real-time monitoring that fires webhooks the moment something new appears. Production agent workflows typically use both.
Reddit, Twitter/X, LinkedIn, Hacker News, GitHub, YouTube, TikTok, Bluesky, DEV, Product Hunt, Stack Overflow, Medium, podcasts, newsletters, and 150k+ news outlets — all through the same endpoint, all pre-scored for relevance and sentiment.
They serve different jobs. The 7-day trial gives you the full product — dashboard, monitoring, alerts — to evaluate Octolens. The sandbox gives you a free API key with a one-time bundle of searches to build and test an integration, with no time limit. When your agent goes to production, plans start at $159/month.
See How Other Teams Use Octolens
Founders
Stay close to your users while you scale. Catch issues early, find leads, and know what your market really thinks.
Learn moreMarketing Teams
Track brand health, measure campaign impact, and monitor competitors — all from one dashboard.
Learn moreDeveloper Relations
Show up where developers talk about you. Respond in minutes, not days. Correct misinformation before it spreads.
Learn moreAgencies
Every client's mentions from one API. Flat pricing, unlimited seats, and data your workflows and agents can build on.
Learn moreGive your agent eyes on the internet
Install the CLI, run octolens signup, and your agent is searching in under a minute — or wire it in over MCP and REST. Never build platform plumbing again.
No credit card required · API access on all plans · Cancel anytime