The AI-visibility CLI,
built for agents

Point a coding agent at Kolaria and it runs the whole loop from the command line: find the questions where AI names a competitor, write the page that wins them back, ship it, and report the lift. One npm install. On every plan.

Start free View the commands Add it to Claude

Get started in seconds

Three lines to your first gap

Install the package, set your token, run the first pass. Your token lives under Settings, then Integrations at kolaria.com.

bash

# install the CLI
npm install -g fulcru

# set your token (Settings -> Integrations)
export FULCRU_TOKEN=pk_...

# find the questions you are losing
$ fulcru gaps

Prefer no account? The free fulcru Claude Skill runs the same playbook with nothing but web access. The CLI is what you reach for when you want it measured, over and over, instead of once.

Runs from any agent

Your agent already knows how to drive it

Point Claude Code, Cursor, Codex, or any MCP-capable agent at the Kolaria CLI and it runs the loop end to end, no dashboard in the way. The output is terse lines, not JSON blobs, because the reader is usually an LLM and its context is the scarce thing.

Claude Code
Cursor
Codex
Devin
OpenCode
n8n
Windsurf
OpenClaw
Hermes

A fulcru gaps call costs an agent about 50 tokens. The same data as raw API JSON costs about 800. Terse output is a design choice, so the agent spends its context on the work.

The loop, in four commands

Find the gap, write the page, ship it, measure the lift

Four commands run the whole play. Add --json to any of them for the raw payload.

01
fulcru gaps
The questions where AI names a competitor and not you, worst first. Pass a number to cap the list.
0% k57jc... best AI visibility tracker -> AI names Profound, Peec AI
02
fulcru write <promptId>
Drafts the page that closes one gap, grounded in what AI actually cites. Markdown to stdout.
fulcru write k57jc... > page.md
03
fulcru publish <pageId> <url>
Marks the page live and snapshots the baseline the next tracking pass measures against.
Published. Baseline: AI names you in 0% of 9 answers.
04
fulcru delta
What your published pages actually did, in mention-rate points. This line is the whole point.
0% -> 44% +44pts Best AI Visibility Trackers, Compared

Want the full picture instead of the gap list? Run fulcru report for overview, competitors, sources, mentions, or prompts.

Claude connector

Add Kolaria to Claude in three steps

Paste one URL and sign in. Claude talks to Kolaria over OAuth, so there is no token to copy and nothing to keep in a config file. It takes about a minute.

01
Open the connector form
In Claude, go to Settings, then Connectors, then Add custom connector.
02
Paste the URL
Use https://api.kolaria.com/mcp. If your client cannot reach that host yet, paste https://little-orca-977.convex.site/mcp instead. Same server, same tools.
03
Sign in and approve
Claude opens a Kolaria sign-in page. Sign in or create an account, approve the request, and the five tools appear in the chat.

What happens the first time

Claude registers itself. Nothing to create beforehand. There is no client id to paste and no key to copy, because the server accepts dynamic client registration.
You sign in once. Approve one screen at kolaria.com. Claude holds the grant and refreshes it on its own, so an hour later the tools still answer without a second sign-in.
Tools read one brand. Every tool works on the brand in your account. A brand-new account should run onboarding at kolaria.com first, or the tools reply that there is no brand to read.

What each tool does

Three read your data. Two change it, and both stay inside the limits your plan already sets, because an agent calls the same guarded code the app calls.

fulcru_gapsReads
List visibility gaps
The questions where AI names a competitor and not you, worst first. Each one carries a promptId for the next tool.
fulcru_write_pageWrites
Write a gap-closing page
Drafts the article that closes one gap, grounded in the sources AI cites in that space. Returns Markdown. Spends the account's AI budget and is rate limited.
fulcru_publish_pageWrites
Publish a page and snapshot baseline
Marks a draft live at a URL and records today's mention rate as the baseline every later pass is measured against.
fulcru_deltaReads
Report published-page delta
For each published page, the mention rate when it went live against the rate now.
fulcru_visibilityReads
Read AI-visibility report
One slice of the brand's report: overview, competitors, sources, mentions or prompts.

OAuth endpoints

You never need these to connect. They are here for anyone who wants to check the flow by hand.

Issuer: https://kolaria.com
Authorize: https://kolaria.com/api/auth/mcp/authorize
Token: https://kolaria.com/api/auth/mcp/token
Register: https://kolaria.com/api/auth/mcp/register
JWKS: https://kolaria.com/api/auth/mcp/jwks
Server metadata: https://kolaria.com/.well-known/oauth-authorization-server
Resource metadata: /.well-known/oauth-protected-resource/mcp, on the MCP host

Token auth, for the CLI

The older path, still working

The CLI and the fulcru-mcp proxy send a token instead of signing in. Anything that cannot do OAuth can use this. Your token lives under Settings, then Integrations at kolaria.com, and it starts with pk_.

Both paths reach the same five tools and the same data. Use the connector above in Claude, and the token here for a shell, a script, or a client with no browser to sign in from.

agent config

{
  "mcpServers": {
    "fulcru": {
      "url": "https://api.kolaria.com/mcp",
      "headers": { "Authorization": "Bearer pk_..." }
    }
  }
}

Endpoint https://api.kolaria.com/mcp. If that host does not resolve for you yet, swap in https://little-orca-977.convex.site/mcp, which is the Convex origin behind it and serves the same server.

Start with a free visibility report

See where AI names a competitor instead of you, no card needed. Then let the CLI or your agent run the loop that closes the gap.

Get your free report Create an account

The AI-visibility CLI, built for agents