Use case

Make your technical docs queryable through MCP

KBrain turns your developer documentation and dev portals into a live, queryable brain - so ChatGPT, Claude, and AI agents always answer from the right version of your docs.

Bring your documentation into ChatGPT and Claude in minutes

Create a docs brain

AI assistants are becoming the first stop for developer questions. The problem is that ChatGPT and Claude were trained on old versions of your documentation - if they were trained on it at all. Developers get confidently wrong answers, file bugs that are not bugs, and waste hours on deprecated patterns. KBrain fixes this by pulling your documentation directly from the source and exposing it as a live, queryable brain through MCP.

The problem: AI assistants answer from stale training data

The Problem - 01

Why AI gives wrong answers about your docs

The model knows your API. Just not the version your developers are shipping against.

👩‍💻

Developer asks

"How do I authenticate with the v3 API?"

›

🧠

Model searches training data

Finds docs from 14 months ago. Picks the most confident answer.

STALE

›

❌

Wrong version returned

v2 pattern. Deprecated token format. Confident, wrong.

VERSION MISMATCH

›

🐛

Developer ships bug

Or wastes 2 hours debugging an AI answer that was never right.

›

🔁

Support ticket opened

Your team answers what the AI should have answered correctly.

The model is not broken. It is working from the knowledge it was trained on. The fix is not a better model - it is a live knowledge source the model can query at runtime.

How KBrain fixes it

KBrain connects to your documentation source - your dev portal, your docs site, your versioned reference pages - and indexes it into a brain. The brain is exposed through MCP. When a developer asks ChatGPT or Claude a question about your API, the assistant queries the brain and returns an answer grounded in your current, live documentation.

The Fix - 02

Live documentation via MCP - how it works

From your docs source to a developer answer in ChatGPT or Claude. No retraining. No manual updates.

❌ Without KBrain

Developer asks ChatGPT about your API.

ChatGPT searches training data. Finds an old version of your docs.

Returns a confident, version-mismatched answer.

Developer implements. Hits a wall. Opens a support ticket.

Your team corrects the answer. Time wasted on both sides.

✅ With KBrain

Developer asks ChatGPT about your API.

ChatGPT queries the KBrain docs brain via MCP.

Brain fetches from your live documentation source. Correct version, current content.

Developer gets an accurate, version-matched answer with a link to the source page.

Implements correctly. No ticket. No wasted time.

KBrain does not retrain the model. It gives the model a live channel to your documentation at query time - so the answer is always grounded in what your docs actually say today.

Set it up in three steps

The Process - 03

From documentation source to queryable brain

No infrastructure to manage. No retraining. No manual sync.

Step 1 - Create a brain

Go to kbrain.io and create a new brain. Name it after your product or documentation set.

Add a description that tells AI agents what this brain contains and when to use it.

Example: "KBrain Docs Brain - contains current API reference, SDK guides, and authentication patterns for v3."

Step 2 - Link your documentation

Connect your docs source: a URL, a docs site, a Google Drive folder, or uploaded files.

KBrain indexes your content. New pages are picked up automatically on the next sync.

Version your brain by documentation version if you maintain multiple active releases.

Step 3 - Expose through MCP

Share your MCP endpoint with your developer community. One URL. Works in Claude, ChatGPT, Cursor, and any MCP-compatible client.

Developers connect once. From that point, every question about your API is answered from your live brain - not from stale training data.

Answers are traceable. Every response includes the source page from your documentation so developers can read the original.

One brain, every AI client. Publish the MCP endpoint in your docs README and every developer who uses an AI assistant gets accurate answers automatically.

Example developer queries

"How do I authenticate with the v3 API? Show me an example request."
"What changed between v2 and v3 of the webhooks endpoint?"
"I am getting a 403 on the /orders endpoint. What permissions does this route require?"
"What is the rate limit for the search endpoint and how do I handle 429 responses?"
"Which SDK methods were deprecated in the latest release and what replaces them?"

Impact - 04

What changes when your docs are live

⚡

version mismatches

Always current

Developers get answers from your current documentation every time. No stale training data. No version confusion.

🎯

✓

source-traced answers

Traceable to source

Every answer includes the source page. Developers can verify and read the original. No guessing whether the AI is right.

🔌

MCP endpoint

Works everywhere

One brain, one endpoint. Works in Claude, ChatGPT, Cursor, and any MCP-compatible client your developers use.

Best-fit documentation types

API reference documentation: endpoint definitions, parameters, request and response formats
SDK and library guides: method signatures, usage examples, migration guides between versions
Authentication and security docs: OAuth flows, API key management, permission scopes
Changelog and release notes: what changed, what was deprecated, what replaces it
Troubleshooting guides: error codes, common failure patterns, resolution steps
Internal tooling docs: platform-specific patterns your developers use every day

You can publish the MCP endpoint directly in your README or developer portal. Any developer who sets it up once gets accurate, live documentation answers in every AI session - without any additional setup from your team.

Bring your documentation into ChatGPT and Claude in minutes

Create a brain, link your documentation source, and expose it through MCP. Developers get accurate, version-matched answers from the first query.

Create a docs brain

Frequently asked questions

Does KBrain stay in sync when we update our documentation?

Yes. KBrain re-indexes your documentation source on a regular sync schedule. When you update a page, add an endpoint, or release a new version, the brain reflects the change on the next sync cycle - without any manual action from your team.

Can we maintain separate brains for different API versions?

Yes. You can create one brain per version - a v2 brain and a v3 brain, for example - each linked to the relevant documentation source. Developers or AI agents connect to the version they need.

Do developers need a KBrain account to query the brain?

No. Developers connect to the MCP endpoint you publish. If the brain is public, they can query it through their AI assistant without a KBrain account. You control whether the brain is public or restricted.

What documentation formats does KBrain support?

KBrain supports web URLs (docs sites, dev portals), PDFs, plain text files, Google Drive folders, and uploaded documents. If your documentation lives on the web or in a file, KBrain can index it.

Can internal documentation be kept private?

Yes. Private brains are accessible only to users or agents you have explicitly shared them with. You can maintain a public brain for your open API reference and a private brain for internal platform documentation.