!Shared memory for coding agents is live.

Memoire
one memory layer,
every coding agent.

Memoire gives Cursor, Claude Code, Codex, and cloud agents one shared context layer for sessions, docs, facts, and project history.

memoire — project: acme-api
$ memoire search "auth middleware refactor"
searching across 12 sessions, 3 clients...
6 results across 3 sessions (847 tokens)
$ memoire timeline --last 3
[cursor] 14:02 refactored auth middleware → jwt validation
[claude] 14:08 tried Redis session cache → reverted
[codex]  14:15 switched to in-memory LRU
$ memoire assemble --budget 4096
assembled prompt block
tokens:
2,847 / 4,096
/ BUILT FOR CROSS-CLIENT ENGINEERING WORKFLOWS:
CursorClaude CodeCodexMCP
/ PRODUCT

Memoire logoMemoire. So every session
picks up where the last left off.

Shared memory, shared docs, shared project facts.
One retrieval loop across every coding surface.

See how Memoire works
session memory
$ memoire collect --session s_a82f
captured 23 events
├─ 8 tool calls
├─ 4 decisions
├─ 2 failed attempts
└─ 9 prompts
$ memoire facts --project acme
fact "using Postgres, not Redis for sessions"
fact "auth uses JWT with 15m expiry"
fact "deploy target: Vercel + Supabase"
cross-client status
Active Clientsproject: acme-api
cursor
last sync: 2m ago
claude code
last sync: just now
codex
last sync: 1h ago
Shared Memory847 events
23
sessions
156
facts
42
docs
version-aware docs
// scanning package.json
"dependencies": {
"next": "15.3.0",
"react": "19.0.0",
"next-auth": "5.0.0"
}
Retrieved docs
next@15.3 App Router, Server Actions, PPR
react@19 use() hook, Actions, Server Components
next-auth@5 Auth.js migration, edge sessions
prompt assembly
Assembling context for: "fix auth bug"
[search]
6 events
[timeline]
3 sessions
[facts]
2 decisions
[docs]
next-auth@5
Prompt blockready
2,847 / 4,096 tokens
/ CAPABILITIES

The context layer
your agents have been missing.

From capture to retrieval to docs search and dashboards, Memoire keeps engineering context in one system.

Shared backend

Store team memory, docs, clients, and retrieval state in one org and project-scoped backend.

backend status
postgresevents: 12,847
pgvectorembeddings: 8,234
collector3 clients
4
orgs
18
projects

Progressive retrieval

Use search, timeline, and assemble to keep prompts compact while still giving agents the full story.

retrieval pipeline
1. search → 24 candidates
2. timeline → 8 recent
3. assemble → 2,847 tokens
compact prompt, full story

Version-aware docs

Memoire indexes the dependency versions in your repo so agents can retrieve current docs instead of guessing from stale training data.

docs index
next@15.3 indexed 847 sections
react@19 indexed 523 sections
prisma@6.1 indexed 312 sections
auto-synced from package.json

MCP + API access

Expose the same memory system through MCP tools, REST APIs, the SDK, and the local dashboard.

access methods
MCP tools stdio / SSE
REST API v1/search, v1/ingest
TypeScript SDK @memoire/sdk
Dashboard trymemoire.com

Deploy your way

Run Memoire locally, self-host it in your own stack, or point clients at a shared hosted backend.

deployment
local memoire start
docker docker compose up
cloud trymemoire.com
/ INFRASTRUCTURE

Built for how teams actually ship

Shared config, project scoping, docs indexing, and deployment options that fit how real engineering teams ship software.

Client-aware installs for Cursor, Claude Code, Codex, and MCP-compatible tools.

PostgresShared

One backend for org, project, event, docs, and client state.

MCP

Local and cloud agents can read and write the same memory surface.

Self-host

Run Memoire in cloud, local, or hybrid environments.

/ USE CASES

Every handoff, every tool switch,
context stays with the work.

Get Started

Use Memoire anywhere you need durable context, exact docs, and shared continuity across sessions and tools.

Session Handoffs

Start a new coding session with prior decisions, failed attempts, touched files, and current project facts already available.

Cross-client Continuity

Move between Cursor, Claude Code, Codex, and cloud agents without losing the thread of the work.

Dependency Guidance

Answer questions about libraries and frameworks with version-matched docs pulled straight into retrieval time.

Team Coordination

Share branch activity, recent changes, and project memory so teammates and agents stop duplicating work.

/ GET STARTED

Stop re-explaining. Start building.

Install Memoire once, connect your project, and let each new session start with shared context instead of a blank prompt.