Feature · SLAtech Business

MCP Server - Your Knowledge Base, Callable as Tools

Your knowledge base is valuable in more places than your website. This is an MCP server over it: one endpoint that speaks the Model Context Protocol, five tools, and a site key that decides which tenant every call belongs to.

Get started free See How It Works

5 tools

Exposed

MCP

Protocol

1 header

To authenticate

120/min

Rate limit per key

Sound Familiar?

🔌

Your Tools Can't Reach Your KB

The knowledge base answers visitors on your site and nowhere else. Internal tooling that could use the same answers has no way in.

🧑‍💻

Manual Copy-Paste Workflow

Someone pastes the docs into another system one file at a time. It works, and it breaks the moment the docs change.

🛠️

One-Off Integrations Add Up

Every consumer that needs the knowledge base gets its own bespoke wrapper. A single documented surface is less to maintain.

What You Get

🔌

A Real MCP Endpoint

POST https://api.slatech.ai/mcp speaks the Model Context Protocol over Streamable HTTP. Ask it for tools/list and it returns each tool with its description and a JSON-Schema for its arguments, so a client configures itself.

🛠️

Five Tools

query_knowledge_base, list_documents, add_knowledge, ingest_url, ingest_sitemap - enough to read the knowledge base, contribute to it, and bring in a whole site at once.

🔑

Site-Key Auth

One header, X-Api-Key. A missing or invalid key is refused with 401 on the transport, before any tool runs. Keys are revocable and auditable.

🏢

No Tenant Argument to Get Wrong

None of the tools take a site identifier. The client and site come from the key alone, so a caller cannot reach another tenant's data by changing an argument - there is no argument to change.

🛡️

Metered and Throttled

Calls count against the same quotas as the rest of your usage, and are rate-limited per key. Writing requires a key carrying the ingest permission; a key issued for the public widget cannot write.

Frequently Asked Questions

What is this, concretely?

An MCP server in front of your knowledge base. A client connects to https://api.slatech.ai/mcp with your site key, calls tools/list to discover the tools and their argument schemas, and then calls them. Discovery is part of the protocol, so there is no separate document to keep in step.

Who's this useful for?

Teams whose internal tooling needs the same answers the website gives - a support console, an onboarding script, a scheduled job that files new documentation. Also agencies running several client knowledge bases from one place, since the key selects which one you are talking to.

Do I need a ready-made connector?

No. It is the protocol itself, not a bespoke API you have to wrap: point a client that speaks MCP at the endpoint and give it the key. If you prefer to call it directly, it is JSON-RPC over HTTP and one header.

How do I index a whole site?

ingest_sitemap takes the URL of an XML sitemap and indexes every page it lists, sitemap indexes included. One call brings the site in, and the crawl runs in the background.

Can add_knowledge or ingest_url be abused?

Both are confined to the site the key owns, and both require a key that carries the ingest permission - a key issued for the public widget cannot write. ingest_url additionally refuses private and loopback addresses, so it cannot be aimed at something internal. Revoking the key stops everything immediately.

Does it count toward my quota?

Yes, on the same meters as everything else. query_knowledge_base counts toward query quota exactly like a website question, and the write tools count toward ingest quota. list_documents reads existing rows and is not metered. MCP calls are also recorded on their own counters, so you can see what the integration costs on its own.

Ready to Get Started?

Free tier. No credit card. Setup in 10 minutes.

Get started free