Agents over your data.
Scoped and on the record.
Build an agent over your own files. It speaks the OpenAI wire format, runs on a credential you can revoke, and every run it makes is recorded as a replayable, auditable trail.
What is our refund policy?
- recallquery: "user prefs"→2 notes29 ms
- searchquery: "refund policy"→4 hits62 ms
- readkey: "pricing-faq.md"→12 KB38 ms
Refunds are processed within 7 business days from the original payment method.
The client you already use.
Swap the base URL, or wrap an existing LangGraph or Vercel AI SDK agent with one import. Either way, the run records itself.
import OpenAI from "openai";
// Already OpenAI-compatible — point the base URL at your agent.
const client = new OpenAI({
baseURL: "https://api.kraterion.com/v1/agents/support",
apiKey: process.env.KRATERION_KEY,
});
const reply = await client.chat.completions.create({
model: "support",
messages: [{ role: "user", content: "What is our refund policy?" }],
});
// → Refunds are processed within 7 business days.
// citations: [pricing-faq.md · §3 · 0.92] · run recordedWhat is our refund policy?
- recallquery: "user prefs"→2 notes29 ms
- searchquery: "refund policy"→4 hits62 ms
- readkey: "pricing-faq.md"→12 KB38 ms
Refunds are processed within 7 business days from the original payment method.
An agent you can
actually govern.
OpenAI-compatible.
Point your existing OpenAI client at an agent's base URL. No new client to learn, no rewrite — your code already works.
Scoped by default.
Each agent gets its own credential, limited to the buckets and tools it needs. Revoke the agent and its access stops at once — no master key, no keys to rotate.
Grounded in your data.
Agents read from your knowledge bases and answer with citations you can verify against the exact source.
On the record.
Pull the full trace of any run, replay it, or audit a single step — retrievals, tool calls, memory, and outputs, all captured.
Tools, ready out of the box.
Every agent ships with tools for searching and reading your data. Add your own, and each call shows up in the run record.
Every run an agent makes.
Recorded and replayable.
Built on the runtime.
Because an agent runs through Kraterion, you get replay, lineage, and memory for free.
Also reachable
from your AI assistant.
Plug Kraterion into Claude Desktop, Cursor, or any MCP-compatible client. Your knowledge, your tools, and persistent memory — over a different protocol, with the same scoped credentials.
{
"mcpServers": {
"kraterion": {
"url": "https://mcp.kraterion.com/mcp",
"auth": {
"type": "oauth",
"dcr": true
}
}
}
}- Claude DesktopFirst-party MCP host
- CursorEditor MCP support
- ZedEditor MCP support
- ContinueIDE assistant
- Any MCP clientImplements the spec
Put your agent on any site.
Issue a share token, paste one script tag, and your customers can ask the agent questions — answered with citations from the buckets you connect. The raw files never leave your project.
<script src="https://embed.kraterion.com/v1.js"
data-token="kr_share_test_..."
defer></script>- Origin-locked tokensEach share token only works on the domains you allow.
- Only cited answers shipThe agent answers from your buckets; the raw files never leave.
- Citations includedEvery reply links back to the source it came from.
Ship an agent
you can stand behind.
OpenAI-compatible. Scoped per agent. Every run on the record.