Documentation

Build agent memory in an afternoon.

Mnemix is the memory and enrichment layer for AI agents. One call before every interaction and your agent knows who it's talking to — designed for sub-300ms voice recall.

Start here

The public API is three endpoints: POST /v1/recall_and_enrich, POST /v1/calls/end, and GET /v1/caller/{phone_number}. Call one when the interaction starts, one when it ends. The guides below walk through installation, the hot path, and the concepts underneath.

Get started

  • Introduction — AI proposes. Mnemix governs.
  • Installation — Mnemix meets your agent where it lives. There are three surfaces, and most teams use more than one:
  • Quickstart — Two parts. Part 1 works right now — live caller memory and enrichment in five minutes. Part 2 is the governance engine in beta — shown so you can see where
  • MCP server — Skip HTTP entirely. The Mnemix MCP server gives any MCP-compatible agent (Claude and others) native memory tools — lookup, save, search, and enrich — witho

Core concepts

  • Overview — Mnemix is built from six primitives. Implementations cycle — the embedder, reranker, storage tiers, and rule engine are all swappable as the field evolves.
  • Evidence refs — Every fact must be backed by evidence. evidencerefs is the primitive that stores provenance: where a claim came from, when the event happened, when Mnemix
  • Locked facts — A locked fact is a value Mnemix surfaces to an agent forever — until a higher-priority source supersedes it. It's the primitive that stops drift: the openi
  • Policy bundles — A policy bundle is a versioned set of rules that codifies a domain's invariants. It's what the gate evaluates a proposed action against. Bundles are how th
  • Validation gate — The gate is the most important piece in Mnemix. It's the barrier between an agent's proposed action and the world. Every high-stakes action passes through
  • Feedback loop — The feedback loop is the axis that lets Mnemix learn from outcomes. A caller agent reports what actually happened after it acted; Mnemix audits the context
  • Bi-temporal validity — Bi-temporal validity is the time model underneath evidencerefs and lockedfacts. It's the foundation of Mnemix's audit guarantee: facts are never deleted or
  • Architecture — How Mnemix is actually built — the worker, the storage tiers, the gate stages, and where the latency budget goes. Every number on this page is an internal