Omem

---

The Problem

Your AI agents have amnesia — and they work alone.

• 🧠 Amnesia — every session starts from zero. Preferences, decisions, context — all gone.
• 🏝️ Silos — your Coder agent can't access what your Writer agent learned.
• 📁 Local lock-in — memory tied to one machine. Switch devices, lose everything.
• 🚫 No sharing — team agents can't share what they know. Every agent re-discovers the same things.
• 🔍 Dumb recall — keyword match only. No semantic understanding, no relevance ranking.
• 🧩 No collective intelligence — even when agents work on the same team, there's no shared knowledge layer.

ourmem fixes all of this.

What is ourmem

ourmem gives AI agents shared persistent memory — across sessions, devices, agents, and teams. One API key reconnects everything.

🌐 Website: ourmem.ai

🧑‍💻 I use AI coding tools

Install the plugin for your platform. Memory works automatically — your agent recalls past context on session start and captures key info on session end.

→ Jump to Quick Start

🔧 I'm building AI products

REST API with 48+ endpoints. Docker one-liner for self-deploy. Embed persistent memory into your own agents and workflows.

→ Jump to Self-Deploy

Core Capabilities

📖 Memory Pipeline Architecture — Technical deep-dive into how ourmem stores, retrieves, and evolves memories.

🔗 Memory Sharing Architecture — How memories flow across agents and teams: sharing, provenance, versioning, and cross-space search.

Feature Overview

Category	Feature	Details
Platforms	4 platforms	OpenCode, Claude Code, OpenClaw, MCP Server
Sharing	Space-based sharing	Personal / Team / Organization with provenance
	Provenance tracking	Every shared memory carries full lineage
	Quality-gated auto-sharing	Rules fire on memory creation (async, non-blocking)
	Vector-enabled shared copies	Shared copies carry source vector embeddings for full search
	Idempotent sharing	Re-sharing returns existing copy (no duplicates)
	Version tracking	Memories track version counter, shared copies detect staleness via ?check_stale=true
	Re-share stale copies	Refresh outdated shared copies with latest source content and vector
	Convenience sharing	One-step cross-user share (share-to-user) and bulk share (share-all-to-user) with auto-bridging
	Organization management	One-step org creation (org/setup) and publish (org/publish) with auto-share rules
	Cross-space search	Search across all accessible spaces at once
Ingestion	Smart dedup	7 decisions: CREATE, MERGE, SKIP, SUPERSEDE, SUPPORT, CONTEXTUALIZE, CONTRADICT
	Noise filter	Regex + vector prototypes + feedback learning
	Admission control	5-dimension scoring gate (utility, confidence, novelty, recency, type prior)
	Dual-stream write	Sync fast path (<50ms) + async LLM extraction
	Post-import intelligence	Batch import → async LLM re-extraction + relation discovery
	Adaptive import strategy	Auto/atomic/section/document — heuristic content type detection
	Content fidelity	Original text preserved, dual-path search (vector + BM25 on source text)
	Cross-reconcile	Discover relations between memories via vector similarity
	Batch self-dedup	LLM deduplicates facts within same import batch
	Privacy protection	<private> tag redaction before storage
Retrieval	11-stage pipeline	Vector + BM25 → RRF → reranker → decay → importance → MMR diversity
	User Profile	Static facts + dynamic context, <100ms
	Retrieval trace	Per-stage explainability (input/output/score/duration)
Lifecycle	Weibull decay	Tier-specific β (Core=0.8, Working=1.0, Peripheral=1.3)
	Three-tier promotion	Peripheral ↔ Working ↔ Core with access-based promotion
	Auto-forgetting	TTL detection for time-sensitive info ("tomorrow", "next week")
Multi-modal	File processing	PDF, image OCR, video transcription, code AST chunking
	GitHub connector	Real-time webhook sync for code, issues, PRs
Deploy	Open source	Apache-2.0
	Self-hostable	Single binary, Docker one-liner, ~$5/month
	musl static build	Zero-dependency binary for any Linux x86_64
	Object storage	AWS S3 or any S3-compatible storage, with IAM role support
	Hosted option	ourmem.ai — nothing to deploy

From Isolated Agents to Collective Intelligence

Most AI memory systems trap knowledge in silos. ourmem's three-tier Space architecture enables knowledge flow across agents and teams — with provenance tracking and quality-gated sharing.

Research shows collaborative memory reduces redundant work by up to 61% — agents stop re-discovering what their teammates already know. — Collaborative Memory, ICLR 2026

	Personal	Team	Organization
Scope	One user, multiple agents	Multiple users	Company-wide
Example	Coder + Writer share preferences	Backend team shares arch decisions	Tech standards, security policies
Access	Owner's agents only	Team members	All org members (read-only)

Provenance-tracked sharing — every shared memory carries its lineage: who shared it, when, and where it came from. Shared copies include the source memory's vector embedding, so they're fully searchable in the target space.

Quality-gated auto-sharing — rules filter by importance, category, and tags. Rules fire automatically when new memories are created. Only high-value insights cross space boundaries.

How It Works

┌──────────────────────────────────────────────────────────────────┐
│   Your AI Agent (OpenCode / Claude Code / OpenClaw / Cursor)     │
│                                                                  │
│   Session Start → auto-recall relevant memories                  │
│   During Work   → keyword detection triggers recall              │
│   Session End   → auto-capture decisions, preferences, facts     │
└───────────────────────────┬──────────────────────────────────────┘
                            │ REST API (X-API-Key)
                            ▼
┌──────────────────────────────────────────────────────────────────┐
│                       ourmem Server                              │
│                                                                  │
│  ┌─ Smart Ingest ─────────────────────────────────────────────┐  │
│  │  Messages → LLM extraction → noise filter → admission      │  │
│  │  → 7-decision reconciliation (CREATE / MERGE / SUPERSEDE / │  │
│  │    SUPPORT / CONTEXTUALIZE / CONTRADICT / SKIP)            │  │
│  │  → cross-reconcile relations → privacy redaction           │  │
│  └────────────────────────────────────────────────────────────┘  │
│                                                                  │
│  ┌─ Hybrid Search (11 stages) ────────────────────────────────┐  │
│  │  Vector + BM25 → RRF fusion → cross-encoder reranker       │  │
│  │  → Weibull decay boost → importance scoring                │  │
│  │  → MMR diversity → parallel cross-space aggregation        │  │
│  └────────────────────────────────────────────────────────────┘  │
│                                                                  │
│  ┌─ Sharing Engine ───────────────────────────────────────────┐  │
│  │  Personal / Team / Organization spaces                     │  │
│  │  → provenance tracking → version-based stale detection     │  │
│  │  → auto-share rules → one-step share-to-user              │  │
│  └────────────────────────────────────────────────────────────┘  │
│                                                                  │
│  ┌─ Lifecycle ────────────────────────────────────────────────┐  │
│  │  Weibull decay (Core β=0.8 / Working β=1.0 / Peripheral   │  │
│  │  β=1.3) → 3-tier promotion → auto-forgetting TTL          │  │
│  └────────────────────────────────────────────────────────────┘  │
└──────────────────────────────────────────────────────────────────┘

• Write once, recall everywhere — memories persist across sessions, devices, and agents
• Gets smarter over time — reconciliation merges, updates, and contradicts memories automatically
• Share across boundaries — Personal → Team → Organization knowledge flow with full provenance
• No manual memory management — Weibull decay handles lifecycle, auto-share rules handle distribution

Key Concepts

API Key = Tenant ID = Your Identity

When you create a tenant, the returned id is your API Key. They're the same UUID. There is no separate "tenant ID".

curl -X POST https://api.ourmem.ai/v1/tenants -d '{"name": "alice"}'
# → {"id": "a1b2c3d4-...", "api_key": "a1b2c3d4-...", "status": "active"}
#          ↑ same value ↑

API Key vs Space ID

Concept	What it is	Example	How many
API Key	Your identity. Goes in X-API-Key header.	a1b2c3d4-...	1 per user
Space ID	A memory storage address. Each is an isolated database.	personal/a1b2c3d4-...	Multiple per user

One API Key owns multiple Spaces:

API Key "a1b2c3d4"
  │
  ├── personal/a1b2c3d4     ← auto-created, your private memories
  ├── team/e5f6g7h8          ← team space you created (you = Admin)
  ├── team/i9j0k1l2          ← team space you were invited to (you = Member)
  └── org/m3n4o5p6           ← organization you joined (you = Reader)

Sharing = Passing Your API Key

To share memories with another user, you pass their API Key as target_user:

# Share all your preferences to user Bob
curl -X POST https://api.ourmem.ai/v1/memories/share-all-to-user \
  -H "X-API-Key: your-api-key" \
  -d '{"target_user": "bobs-api-key", "filters": {"categories": ["preference"]}}'

The system automatically creates a bridging Team Space between you and Bob. No manual space management needed.

Quick Start

Agent Install (recommended)

One message to your AI agent. It handles everything — API key, plugin install, config, verification.

Hosted (ourmem.ai — nothing to deploy):

Platform	Copy this to your agent
OpenClaw	Read https://ourmem.ai/SKILL.md and follow the instructions to install and configure ourmem for OpenClaw
Claude Code	Read https://ourmem.ai/SKILL.md and follow the instructions to install and configure ourmem for Claude Code
OpenCode	Read https://ourmem.ai/SKILL.md and follow the instructions to install and configure ourmem for OpenCode
Cursor / VS Code	Read https://ourmem.ai/SKILL.md and follow the instructions to install and configure ourmem as MCP Server

Self-hosted (your own server):

Platform	How to install
OpenClaw	Run openclaw skills install ourmem, then tell your agent: setup ourmem in self-hosted mode
Claude Code	Read https://raw.githubusercontent.com/ourmem/omem/main/skills/ourmem/SKILL.md and install ourmem for Claude Code, self-hosted mode
OpenCode	Read https://raw.githubusercontent.com/ourmem/omem/main/skills/ourmem/SKILL.md and install ourmem for OpenCode, self-hosted mode

That's it. Your agent handles the rest.

Skill Install (alternative):

If you prefer CLI installation, install the ourmem skill directly:

npx skills add ourmem/omem --skill ourmem -g

This works with 44+ AI agents including Claude Code, OpenCode, Cursor, and more. See Vercel Skills CLI for details.

---

1. Get an API Key

Hosted:

curl -sX POST https://api.ourmem.ai/v1/tenants \
  -H "Content-Type: application/json" \
  -d '{"name": "my-workspace"}' | jq .
# → {"id": "xxx", "api_key": "xxx", "status": "active"}

Self-deploy:

docker run -d -p 8080:8080 -e OMEM_EMBED_PROVIDER=bedrock ghcr.io/ourmem/omem-server:latest
curl -sX POST http://localhost:8080/v1/tenants \
  -H "Content-Type: application/json" \
  -d '{"name": "my-workspace"}' | jq .

Save the returned api_key — this reconnects you to the same memory from any machine.

2. Install Plugin

OpenCode: Add "plugin": ["@ourmem/opencode"] to opencode.json + configure plugin_config with apiUrl and apiKey in the same file.

Claude Code: /plugin marketplace add ourmem/omem + set env vars in ~/.claude/settings.json.

OpenClaw: openclaw plugins install @ourmem/ourmem + configure openclaw.json with apiUrl and apiKey.

MCP (Cursor / VS Code / Claude Desktop):

{
  "mcpServers": {
    "ourmem": {
      "command": "npx",
      "args": ["-y", "@ourmem/mcp"],
      "env": {
        "OMEM_API_URL": "https://api.ourmem.ai",
        "OMEM_API_KEY": "your-api-key"
      }
    }
  }
}

3. Verify

curl -sX POST "$OMEM_API_URL/v1/memories" \
  -H "X-API-Key: $OMEM_API_KEY" -H "Content-Type: application/json" \
  -d '{"content": "I prefer dark mode", "tags": ["preference"]}'

curl -s "$OMEM_API_URL/v1/memories/search?q=dark+mode" -H "X-API-Key: $OMEM_API_KEY"

What Your Agent Gets

Memory Tools

Tool	Purpose
memory_store	Save facts, decisions, preferences with smart dedup
memory_search	11-stage hybrid search (vector + BM25 + reranker)
memory_get	Retrieve a specific memory by ID
memory_update	Modify content, tags, importance, tier
memory_delete	Remove a memory
memory_list	Browse memories with filters and pagination
memory_ingest	Smart-ingest full conversations (LLM extraction)
memory_profile	Auto-generated user profile (static facts + dynamic context)
memory_stats	Analytics: tag distribution, decay curves, relation graphs

Sharing Tools

Tool	Purpose
space_create	Create Team or Organization spaces
space_list	List all accessible spaces
space_add_member	Invite users to a shared space
memory_share	Share a memory to any space with provenance
memory_pull	Pull a shared memory into your personal space
memory_reshare	Refresh stale shared copies with latest content

Convenience APIs: share-to-user (one-step cross-user share with auto-bridging) and share-all-to-user (bulk share). MCP exposes 15 tools + 1 resource. OpenCode/OpenClaw expose 11 tools + 3 hooks.

Smart Hooks

Hook	Trigger	What Happens
SessionStart	New session begins	Recent memories + user profile auto-injected into context
KeywordRecall	Keyword detected mid-session	Relevant memories surfaced without explicit search
SessionEnd	Session ends	Key decisions, preferences, and facts auto-captured
Compaction	Context window compressed	Critical memories preserved across compaction boundaries

Memory Space

Browse, search, and manage your agent's memories visually at ourmem.ai/space — see how memories connect, evolve, and decay over time.

Security & Privacy

Rust Memory Safety	No garbage collector, no data races. Ownership model guarantees safety at compile time.
Tenant Isolation	X-API-Key auth with query-level tenant filtering. Every operation verifies ownership.
Privacy Protection	<private> tag redaction strips sensitive content before storage.
Encryption	HTTPS for all API transit. Server-side encryption at rest on S3.
Admission Control	5-dimension scoring gate rejects low-quality data before storage.
Open Source Auditable	Apache-2.0 licensed. Audit every line, fork it, run your own instance.

Self-Deploy

# Minimal (BM25 search only, no embedding API needed)
docker run -d -p 8080:8080 ghcr.io/ourmem/omem-server:latest

# With Bedrock embedding (recommended, needs AWS credentials)
docker run -d -p 8080:8080 \
  -e OMEM_EMBED_PROVIDER=bedrock \
  -e AWS_REGION=us-east-1 \
  ghcr.io/ourmem/omem-server:latest

# With OpenAI-compatible embedding
docker run -d -p 8080:8080 \
  -e OMEM_EMBED_PROVIDER=openai-compatible \
  -e OMEM_EMBED_API_KEY=sk-xxx \
  ghcr.io/ourmem/omem-server:latest

Full deployment guide: docs/DEPLOY.md

Object Storage (Optional)

By default ourmem stores data on local disk. For durability and scalability, configure AWS S3 or any S3-compatible object storage:

# Add to your environment
OMEM_S3_BUCKET=your-bucket            # enables s3:// scheme
AWS_ENDPOINT_URL=https://s3.amazonaws.com
AWS_REGION=us-east-1

# Alternative: OSS-compatible storage
OMEM_OSS_BUCKET=your-bucket           # enables oss:// scheme
OSS_ENDPOINT=https://oss-xx-internal.aliyuncs.com
OSS_ACCESS_KEY_ID=your-ak             # or use ECS RAM role (auto-discovered)
OSS_ACCESS_KEY_SECRET=your-sk

If both OMEM_S3_BUCKET and OMEM_OSS_BUCKET are set, OSS takes priority.

Build from Source

Two build modes

Mode	Command	Binary	Bedrock	Runs on
glibc (full)	cargo build --release	Dynamic linked, ~218MB	✅ AWS Bedrock	Same glibc version as build host
musl (portable)	See below	Static linked, ~182MB	❌ OpenAI-compatible only	Any Linux x86_64

glibc build (with Bedrock support)

cargo build --release -p omem-server
# Binary: target/release/omem-server
# Requires: same or newer glibc on target machine

musl static build (portable, zero dependencies)

Single binary that runs on any Linux x86_64 — no glibc, no libraries, nothing.

rustup target add x86_64-unknown-linux-musl

RUSTFLAGS="-C target-feature=+crt-static -C relocation-model=static" \
  cargo build --release --target x86_64-unknown-linux-musl \
  -p omem-server --no-default-features

# Binary: target/x86_64-unknown-linux-musl/release/omem-server
# Statically linked, runs anywhere

Note: The musl build uses --no-default-features which excludes AWS Bedrock support. Use OMEM_EMBED_PROVIDER=openai-compatible (e.g. DashScope, OpenAI) instead. This is because aws-lc-sys (AWS crypto library) crashes on musl static linking due to dlopen(NULL) incompatibility (aws-c-cal#213), and Rust's default static-pie output segfaults with musl-gcc (rust-lang/rust#95926).

Transfer to any server

# Compress
gzip -c target/x86_64-unknown-linux-musl/release/omem-server > omem-server.gz

# Copy to server
scp omem-server.gz user@server:/opt/

# Run (no dependencies needed)
ssh user@server "gunzip /opt/omem-server.gz && chmod +x /opt/omem-server && /opt/omem-server"

API at a Glance

Method	Endpoint	Description
POST	/v1/tenants	Create workspace & get API key
POST	/v1/memories	Store memory or smart-ingest conversation
GET	/v1/memories/search	11-stage hybrid search
GET	/v1/memories	List with filters & pagination
GET	/v1/profile	Auto-generated user profile
POST	/v1/spaces	Create shared space
POST	/v1/memories/:id/share	Share memory to a space
POST	/v1/files	Upload PDF / image / video / code
GET	/v1/stats	Analytics & insights

Full API reference (48+ endpoints): docs/API.md

Documentation

Document	Description
docs/API.md	Complete REST API reference
docs/PIPELINE.md	Memory pipeline architecture — storage, retrieval, and plugin integration flows
docs/SHARING.md	Memory sharing architecture, flows, and tutorials
docs/DEPLOY.md	Docker & AWS deployment guide
docs/PLUGINS.md	Plugin installation for all 4 platforms
skills/ourmem/SKILL.md	AI agent onboarding skill

License

Apache-2.0

---