Documentation

Create an account — free workspaces start with 30 credits.

Connect a source you control, or run an ad-hoc analysis from the dashboard terminal.

Watch the item move through extraction, analysis and indexing, then open it to read its semantic index components.

Create a test API key and call /v1/content to consume the same data from your own system.

A source represents content you own or are authorized to index. You configure the platform, the URL or handle, a category and a refresh interval.

Source status always reflects the last real check. A source that has never been checked shows as pending, not connected.

When a re-check detects a change, the semantic index is recomputed and a new provenance version is written. Earlier versions are retained.

The index is a weighted 0–100 score over seven components. It measures metadata structure and coverage — not popularity, and not ranking potential.

Every component value is stored with the item so a score can always be explained.

Commands are plain text. `help` lists every command, Tab completes the highlighted suggestion, and ↑/↓ recall history.

Inside a signed-in workspace, anything phrased as a question is routed to Copilot, which answers strictly from your own indexed data and says so when the answer isn't present.

Content bytes and canonical metadata JSON are hashed separately with SHA-256, then combined with the timestamp and source into a provenance hash.

Records are publicly readable by hash or record ID at /verify. Hashes only — never content, never personal data.

Credits are reserved atomically before any analysis runs, so balances can never go negative and a failed reservation performs no work.

Rate limits are per plan and per minute. Exceeding them returns 429 with a Retry-After header.