
Fulcru
io.github.gsmmediaroPublished by GitHub account gsmmediarov2.0.0Updated Oct 25, 2026
Find where ChatGPT names your competitor, write the page that closes the gap, measure it.
Overview
Fulcru is a hosted MCP server that finds AI-search questions where competitors are named instead of your brand, drafts the page that closes the gap, and tracks
- What you can do
- Find prompts where AI assistants name competitors instead of your brand, worst first, and draft a Markdown page that closes one gap. Mark a drafted page live with its URL and snapshot a baseline to measure from. Report what published pages changed in AI visibility, plus overview, competitor, source and mention reports.
- When to use it
- Use it when you want an assistant to run an AI-search visibility loop: audit which questions name competitors, produce gap-closing content, and measure the before/after change. It suits teams tracking brand mentions across AI assistants continuously rather than one-off checks.
- Requirements
- A remote endpoint reached over streamable HTTP, authenticated with a bearer token in the Authorization header obtained from the provider's Integrations page; OAuth 2.1 with dynamic client registration is also supported. A Fulcru account is needed for the measured loop, and network access to the hosted service.
Installation
In SourceWeft
- Open Fulcru in the dashboard and add it to a workspace.
- Connect Authorization token credentials. They are stored privately and never shown on public pages.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Use with an agent
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.gsmmediaro%2Ffulcru/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Other MCP clients
Add this to your client's mcpServers config.
{
"mcpServers": {
"fulcru": {
"type": "http",
"url": "https://api.fulcru.app/mcp",
"headers": {
"Authorization": "Bearer <YOUR_TOKEN>"
}
}
}
}README
fulcru
Get named by AI search engines.
When someone asks ChatGPT "what's the best CRM for a small team", the answer names two or three products. Those get the customer. Everyone else is invisible.
fulcru finds the questions where an AI assistant names a competitor instead of
you, writes the page that closes the gap, and measures what publishing it did. It
runs as a CLI, an MCP server, and an agent skill, so a human or an agent can run
the whole loop.
This repo is the open, self-hostable agent surface. The measurement engine is the hosted Fulcru service; everything here is a thin client that talks to it with your token.
We run it on ourselves and publish the result, whichever way it goes. Today it scores us 0 out of 34 prompts: we sell AI visibility and we score zero on our own product, with the full run and the source data.
Or run it in Docker, with nothing installed on the host:
No secrets are baked into the image; the token is passed at run time. Verified on
node:20-alpine, 2026-07-27.
The loop
That last line is the whole point.
Commands
Add --json to any command for the raw payload.
Built for agents
Output is terse lines, not JSON blobs, because the caller is usually an LLM and
its context is the scarce resource. fulcru gaps costs an agent about 50 tokens;
the same data as raw API JSON costs about 800. fulcru write sends the Markdown
to stdout and everything else to stderr, so fulcru write <id> > page.md gives a
clean file.
MCP
Two ways to wire the five Fulcru tools into an agent.
Remote (hosted). Point any MCP client at the hosted endpoint with your token:
Local (stdio). For hosts that prefer a subprocess, this package ships a stdio MCP server that proxies to the hosted endpoint:
The -p is not decoration. This package ships two commands, and without it npx
runs the one named after the package, which is the CLI: the host gets a usage
screen instead of an MCP server.
Tools: fulcru_gaps, fulcru_write_page, fulcru_publish_page, fulcru_delta,
fulcru_visibility.
Both speak protocol revision 2026-07-28 and the older handshake-based
revisions, and pick per request from what the client sends. A client on the new
revision gets server/discover, per-request _meta, and a cacheable tool list;
one still sending initialize is served exactly as before. Nothing needs
configuring either way.
Skill
The skills/fulcru skill runs the whole playbook using
nothing but web access — audit a domain, find the gaps, write the page, re-measure
— with no account and no token. Install it into an agent that supports skills:
Hermes scans third-party skills before installing and prints the verdict, so you can see what you are agreeing to run. If you would rather track this repo as a source and pick up updates, tap it instead:
The CLI and MCP above are what you use when you want it measured continuously across every engine, with the before/after delta tracked for you.
Configuration
No secrets are baked into this repo. The token is read from the environment only.
Self-hosting
The skill needs nothing. The CLI and MCP need a Fulcru account for the continuous,
measured loop; point FULCRU_ENDPOINT at your own instance if you run one.
A Docker image is provided:
Contributing
Issues and PRs welcome. The tool logic lives in the hosted service; this repo is the CLI, the stdio MCP proxy, and the skill. If AI names a competitor instead of you, open an issue with the question — that is exactly the gap this is built to close.
If this is useful, a ⭐ genuinely helps.
License
AGPL-3.0-or-later. See LICENSE.
Source: README.md at commit 56903ae
Tools
0Version history
1- v2.0.0LatestSep 16, 2026
Score
85% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
- Auth declaredRequired10 pts
- Tools listedRequired15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts