Everything Obi does

Learns your business

Reads your sources

Obi reads your site and blog, third-party reviews and "alternatives" pages, then the home, product and pricing pages of your rivals.

Renders what visitors see

Obi opens JavaScript pages, price toggles and visuals in your agent's browser, so a fact hidden from a plain read still counts.

Watches your videos

Obi saves a video's details, its captions and up to 30 still frames. Run obi research video.

Keeps every source

Every page Obi reads is saved with its address, the time it was read and the passages quoted. A page that failed stays listed as failed.

Finds your ICP

Obi proposes your offer, buyer, market and competitors, each claim cited. When your site and others disagree, Obi picks the buyer.

Takes your corrections

You correct in plain words: fix a name, fill a fact, dispute or replace a claim, add or remove a competitor. Run obi business.

Writes your buyers' questions

From the buyer's seat

Obi writes each question as your buyer types it into an AI: first person, one ask, a real situation, the fact it came from.

The whole buying path

The set covers every stage of awareness and six intents: category, use case, feature, comparison, alternatives and trust.

Any country and language

Obi works in any country and language: it asks the AI engines a startup's buying questions in the country and language where the startup sells.

You review the set

Obi shows five representative questions, then all of them. Your wording becomes a new version, and every scan records the version it asked. Run obi questions show.

Scans AI answers

On Obi's servers

Obi's servers ask the engines your questions. A scan is the only thing Obi calls a measurement; your agent's own answer stays a suggestion.

Ten engines

Obi scans ChatGPT, Gemini and Google AI Overviews in every core pack, with Claude and Perplexity as options. Obi also scans Grok, Microsoft Copilot, Meta AI, DeepSeek and Mistral Le Chat.

Answer, date, sources

Each result keeps what the engine answered, when, and the sources it cited. Run obi questions results.

Every week

A core pack is scanned every week; the Free plan once a month. A scan freezes its questions when it starts, so a later edit never rewrites it.

Your yes first

Obi shows the questions, the engines and the count before a scan, and only your explicit yes starts it.

Checks your website

Crawler access

Obi reads your robots.txt rule by rule for each AI search crawler and each page, and names the line that decides.

Indexability

Obi finds pages that are down, marked noindex or nosnippet, restricted from Bing answers, or pointing their canonical elsewhere.

JavaScript pages

Obi finds the facts that appear only once a browser runs the page, which AI crawlers may never see.

Structured data and llms.txt

Obi checks your structured data and whether your llms.txt exists, is valid and links to pages that answer.

Every finding sourced

Each finding carries a severity, a confidence, the quoted evidence and how to fix and verify it. A robots.txt fix comes as an exact patch.

Training crawlers apart

Blocking a training crawler such as GPTBot is reported as your choice. Obi keeps it apart from the crawlers that power AI answers. Run obi readiness.

Does the work

Three actions this week

After a scan, Obi ranks what to do by urgency and buyer importance, then evidence, speed and effort. Each action names the observation it rests on. Run obi strategies.

The considered list

Obi shows what else it weighed and why it ranks lower, and says what would produce new ideas when the list runs out.

Drafts in your voice

Obi writes the missing pieces: pages, articles, comparisons, FAQs, fix briefs, pitches and replies.

Every draft saved

Obi saves each draft with its revisions, so you can correct, resume or export it. Run obi harness artifact.

Records what you shipped

When you make a change, Obi records what changed, when and the evidence, and reads it before its next decision.

Agent Search

Agent Search is Obi's work for AI agents: Obi asks coding agents and personal agents what a startup's buyers ask them, sees whether they pick the startup, and writes what they read before they choose (docs, setup steps, product facts and prices), so they can install it or buy it.

Works with your tools

Attio

Obi reads the won deals to find the ICP, and ties the leads AI brings to revenue.

Granola

Obi reads the sales calls recorded there, the buyers' real questions and objections.

GitHub

Obi opens pull requests on the product's repository for the pages and docs it writes.

Mintlify

Obi publishes the product's docs ready for agents, as Markdown pages with an llms.txt file.

Google Search Console

Obi reads the queries, the indexing, and the impressions in Google's AI features.

G2

Obi reads the reviews, the buyers' own words and a source AI engines cite.

Reddit

Obi reads the threads buyers trust, a source AI engines cite.

YouTube

Obi reads the demos and their transcripts, where buyers learn.

LinkedIn

Obi reads the buyers' and the company's own words.

PostHog

Obi shows the visits and sign-ups that come from AI.

Runs in your agent

Your coding agent

Obi runs in Claude Code, Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI, Cline, Amp and OpenCode, or any terminal. Type /obi in Claude Code or Cursor, $obi in Codex.

Your agent's power

Obi researches and reasons with your agent's own search, browser and subagents, reading your site, the reviews and your rivals in parallel.

Remembers you

Obi keeps the preferences, rules and decisions you ask it to remember, across agents and sessions. Run obi memory.

Speaks your language

Obi follows the language of your conversation; the commands, sources and saved questions stay exact.

Picks up where it stopped

Obi resumes an interrupted job from its saved state, and reads before it retries a write, so a paid scan is never sent twice.

Built for scripts

Every command answers in one validated JSON envelope with --json; obi doctor checks your setup without sending anything.

Stays in your hands

Nothing goes live alone

Obi recommends and drafts. It never publishes, sends outreach or buys anything on its own.

Paid work needs your yes

Every command that starts paid work or approves your review takes an explicit confirmation, so it never runs by accident.

Your connections

Obi keeps its sign-in in your system's credential store and never prints a token. You list or revoke each machine. Run obi credentials.

Plans in your agent

Obi quotes the packs for the questions you reviewed and opens the checkout only after your confirmation. Free: 10 questions; each core pack: $20 a month for 25 questions.

Planned

Obi's API and Obi's MCP server are planned. Today Obi works through its CLI and its Skill.