Everything below authenticates with your API tokens (create them on the Tokens tab). Always include the manufacturer in queries — it measurably improves identification.
Quick start — REST
One call: messy description in, canonical record out. Unknown products queue background research when your token has the research scope (spends your monthly quota).
Prefer markdown (for LLM prompts and RPA logs)? Append .md to any product URL or send Accept: text/markdown:
Batches — submit a manifest's worth at once, then poll (or register a webhook with callbackUrl):
MCP — connect your AI agents
Katalog ships a built-in Model Context Protocol server. Point Claude (Desktop, Code, or API) or any MCP client at it and your agents can resolve asset strings, read records, track research, and file accuracy feedback — governed by the same token scopes and research quota as the REST API.
Claude Desktop / Claude Code configuration:
| Tool | Scope | What it does |
|---|---|---|
lookup_product | lookup | Resolve a raw description to a record; queues research for unknowns (uses quota). |
get_product | products | Fetch a canonical record as markdown by productId. |
research_status | lookup | Poll a queued research job. |
submit_feedback | products | Report inaccurate data — lands in our human review queue. |
A typical agent flow: lookup_product("HP E45028DN LASERJET") → record markdown, or a jobId → research_status(jobId) a minute later → get_product(productId). If the agent spots a wrong value: submit_feedback.
Full API reference
Try a lookup
Runs the matcher exactly as the API would — read-only: no alias learning, no research, no quota use.
Request research
Not in the catalog? Submit it for research.
Your research jobs
| Submitted | Query | State | Result |
|---|
Your accuracy reports
| When | Product | Field | Message | Status |
|---|
| Label | Token | Scopes | Created | Last used | Status |
|---|
| Day | Metric | Count |
|---|