Truememory Developer Docs
Explore guides and API reference to start building with Truememory
Powerful primitives, seamlessly integrated
Sign users in with one call
Email/password and OAuth (Google, Apple) through Truememory Auth, built on Google Identity Platform. Every request to Truememory carries the Bearer token you get back.
curl -X POST $TRUEMEMORY_AUTH_URL/auth/loginThe OpenAI-compatible gateway
The same /v1/chat/completions endpoint every OpenAI SDK already knows — with memory, personas, model routing, and observability added server-side.
curl -X POST $TRUEMEMORY_API_URL/v1/chat/completionsMemory that follows your users
Persistent user memory on a Graphiti knowledge graph, injected into every conversation automatically — searchable, mergeable, and scoped per app.
curl -X POST $TRUEMEMORY_API_URL/memories/searchGround responses in curated knowledge
Bundle knowledge packs into brains and attach a whole collection to a chat request with a single brain_id.
curl -X POST $TRUEMEMORY_API_URL/brains/$BRAIN_ID/knowledge-packsBuild with your favorite AI agent
Paste into any AI coding agent to point it at the Truememory docs index — every page ships a markdown twin agents can read directly:
What's new
The latest features and improvements shipping across Truememory.
Memory graph: one graph, every entity visible
GET /memories/graph no longer splits episodes into a notes layer or hides entities that only an API ingest mentioned, and a `document` ingest type replaces `note`.
Id-aware record resolution for shared graphs
Member records keep their identity — same-name members stay distinct, corrections find the right node, and id references become edges.
Graph namespaces: full entity graph on inspection, service token required
GET /memories/graph with a graph_id now returns the namespace's deduplicated entity graph instead of an empty entity layer, and requires a service token.
Brains: group knowledge packs into one collection
A brain bundles knowledge packs under one ID, so a chat request can attach the whole set with brain_id.
Graph namespacing — blend shared knowledge graphs into memory search
New graph_ids field on POST /memories/retrieve and POST /memories/search blends named, user-independent knowledge graphs into memory retrieval concurrently with the user's own memory, with per-result provenance.
Guides for every step
End-to-end recipes and the authentication setup that everything else builds on.
Cookbooks
Collect user identity through one onboarding agent
The unified onboarding agent gathers identity fields conversationally and stores them for every app.
Run onboardingOne agent per user, shared across all your apps
Cross-app agent setup — create once, select per request with the X-Agent-Id header.
Create an agentWire persistent memory into your product
Integrate memory end to end: writes, searches, timelines, and per-app graph scoping.
Integrate memoryWays to authenticate
Pick the right credential for every caller
User tokens for people, service accounts for backends — how to choose and what each unlocks.
Compare methodsSign real people in from web and mobile
Email/password and Google/Apple OAuth via Truememory Auth, with refresh flows and email verification.
Set up sign-inCall Truememory from your backend services
Webhooks, cron jobs, and microservices authenticate with a shared AUTH_SECRET — no sign-in flow.
Use a service accountEverything behind one endpoint
Audio, images, voice, threads, personas, and skills — all on the same OpenAI-compatible API surface.
Speech-to-text and text-to-speech in two calls
OpenAI-compatible /v1/audio/transcriptions and /v1/audio/speech endpoints.
Add audioGenerate images with OpenAI and Gemini image models
The /v1/images/generations endpoint routes to OpenAI and Google Gemini image models.
Generate imagesGive personas a voice with per-user preferences
Real-time voice sessions with voice preferences stored per persona.
Enable voiceConversation state that survives requests
List, rename, and delete conversation threads with full history.
Manage threadsA custom system prompt for every user
Create, activate, and manage per-user personas — injected automatically on chat.
Create a personaReusable capabilities, scoped per app
Attach skills to agents with multi-app scoping and category management.
Browse skillsConnect with Truememory
Browse the source, follow releases, and point your agents at the docs.
Follow what ships
New features, improvements, and fixes across Truememory Auth and the AI Gateway.
Read the docs like a machine
Every page ships a markdown twin, plus site-wide indexes for AI agents.