Documentation Index
Fetch the complete documentation index at: https://docs.complyhat.ai/llms.txt
Use this file to discover all available pages before exploring further.
What compliance memory is
Compliance memory is a single master markdown document per organization: the framework interpretations, methodology choices, model contexts, policy excerpts, and findings your AI agent recalls when drafting future reports. It is the product’s memory — every approved report compounds into it, so each new document inherits the positions you already took. It replaces the older multi-page wiki: instead of many hyperlinked pages with per-chunk embeddings, there is one document your host reads whole into its context window. Smaller surface, fewer moving parts, no embeddings to manage. Two things feed it:- Approved reports append to it automatically. When the host calls
reports.finalize, ComplyHat appends an approved-positions block (report name, frameworks, models, content hash) to the document and returns the newmemory_version. No host follow-up call is needed. - Host edits — your agent can read the document, edit it, and write it back (
wiki.write), or append a note to the end (wiki.append). Use this to record methodology decisions, threshold rationale, regulator correspondence, or to fold in an internal policy your agent has read locally.
How your agent uses it
On everyreports.start_draft, ComplyHat returns the whole compliance-memory document alongside the template and fresh evidence. The host reads it to stay consistent with prior approved positions — same thresholds, same framework interpretations, same methodology — instead of re-deriving them each time.
To edit it, your agent calls wiki.read (returns the markdown plus a version), makes its changes, and calls wiki.write with the full document. Passing prev_version enables optimistic concurrency: if the document changed underneath you, the write returns 409 and you re-read. For a quick addition, wiki.append adds a block to the end without re-sending the whole file. wiki.delete clears the whole document if you need to start over.
ComplyHat never generates this text. The host composes it; ComplyHat stores markdown only — the zero-LLM invariant.
Ingesting a document
To fold an internal policy, regulator letter, or vendor report into your compliance memory, tell your agent: “Read this PDF and add the key points to our compliance memory.” The host reads the file locally, extracts the relevant text, and appends it viawiki.append (or merges it with wiki.write). The original file never leaves your machine; ComplyHat persists only the markdown your host sends.
Reading, exporting, and clearing
The dashboard Report Library has a Compliance memory (.md) download that returns the whole document as a Markdown file. Your agent can also pull the document any time withwiki.read.
The dashboard is read-only with one exception: a Delete control next to that download clears your whole compliance memory. Your agent can do the same with wiki.delete, which empties the document while keeping its version history, for the customer who wants to erase their accumulated posture. To add or edit, ask your MCP host.