Everything in the box
One install gets you the engine, the MCP server, the dashboard, and the Python library. Here's exactly what's inside, how to install it, and what each tier unlocks.
Your agent's memory, as a living map you can grab
Every memory becomes a linked entity. Open the Graph tab and the whole knowledge base lays itself out in front of you — hundreds of nodes you can reach in and move.
Real-time physics, hundreds of nodes
A force-directed layout runs live in your browser. Grab a node and the graph re-settles instantly — 345 memories here, no lag and no round-trip to a server.
A tool, not a picture
Drag nodes to pull threads apart; tune repel force, link distance and gravity; search for any entity; then Refresh, Reheat, or Fit. Hide unconnected noise or reveal every link label.
See the shape of what it knows
Related memories link together and the most-connected concepts rise to the top, so the themes your agent keeps returning to are obvious at a glance.
Pick the install that matches what you need
First run downloads all-MiniLM-L6-v2 (~80 MB). Without it, the engine falls back to a deterministic offline embedder. It always runs.
Three ways to run it
The headline experience
Claude Code, Cursor, Zed
Same service, no server
18 tools, once your agent connects
engraphis_remember and engraphis_recall, scoped to workspace, repo, and session, plus governance, code awareness, and session lifecycle.
| Category | Tool | What it does |
|---|---|---|
| Write | engraphis_remember | Store a fact; deterministically resolved (add / reinforce / supersede) |
| Write | engraphis_record_event | Append a lightweight episodic log entry |
| Write | engraphis_link | Explicitly connect two related memories |
| Read | engraphis_recall | Hybrid vector + lexical + graph recall |
| Read | engraphis_recall_grounded | Cited answer from retrieved memories, or abstain |
| Read | engraphis_recall_proactive | "What should I know right now," no query needed |
| Read | engraphis_why | Current answer, plus what it superseded |
| Read | engraphis_timeline | Full bi-temporal history, oldest first |
| Code | engraphis_index_repo | Parse a repo into the code symbol graph |
| Code | engraphis_search_code | Find symbols by name, with callers |
| Governance | engraphis_forget | Retire a memory: bi-temporal close, never deleted |
| Governance | engraphis_pin | Exempt from future automatic decay / pruning |
| Governance | engraphis_correct | Replace content without losing history |
| Session | engraphis_start_session / _end_session | Session lifecycle with cross-session handoff |
| Ops | engraphis_stats | Memory counts for health checks |
Set entirely through environment variables
Set these directly or drop them in a .env file. See .env.example in the repo for the full list.
| Env var | Default | Description |
|---|---|---|
ENGRAPHIS_DB_PATH | ./engraphis.db | SQLite database file |
ENGRAPHIS_HOST | 127.0.0.1 | Server bind address |
ENGRAPHIS_PORT | 8700 | Dashboard port |
ENGRAPHIS_API_TOKEN | — | If set, REST API requires a bearer token |
ENGRAPHIS_EMBED_MODEL | all-MiniLM-L6-v2 | sentence-transformers model |
ENGRAPHIS_LLM_PROVIDER | openai | openai | anthropic | google | openrouter |
ENGRAPHIS_LLM_API_KEY | — | LLM API key (only for optional chat/synthesis) |
ENGRAPHIS_LICENSE_KEY | — | Pro/Team key (or ~/.engraphis/license.key) |
How Engraphis stacks up
Engraphis owns self-hosted, coding-agent-native, and no per-operation metering. Where a competitor is the better pick, we say so.
vs. mem0
mem0 is an Apache-2.0 memory library with a hosted SaaS on top: a genuinely good product. The difference is the deployment model and where the paywall falls. mem0 gates graph memory behind a paid tier; Engraphis keeps the whole recall engine free.
| mem0 | Engraphis | |
|---|---|---|
| Default deployment | Hosted-first SaaS | Self-hosted-first, local by default |
| Billed per memory / retrieval | Yes, hosted | No, runs on your box |
| Graph memory | Gated to a paid tier | In the free core |
| Coding-agent integration | SDK glue | Native MCP server (18 tools) |
| Code symbol graph | No | Yes |
vs. Zep
Zep is a cloud, credit-metered memory service that deprecated its self-hosted Community Edition, vacating the on-prem/private niche entirely. Any team disqualified from a hosted memory cloud is a customer Zep has walked away from.
| Zep | Engraphis | |
|---|---|---|
| Deployment | Cloud only, self-hosted CE deprecated | Self-hosted, first-class |
| Billing model | Credit-metered per operation | No metering, flat license |
| Offline / air-gapped | No | Yes |
| On-prem / regulated / residency-bound | No longer served | The core use case |
vs. Letta (MemGPT)
Letta is an open-source agent framework with memory built in, and its free tier is genuinely generous. The real difference is scope: Letta asks you to build on it; Engraphis asks you to add it.
| Letta | Engraphis | |
|---|---|---|
| What it is | Agent framework with memory inside | Drop-in memory layer, bring your own agent |
| Adoption cost | Adopt the framework | Add one MCP server, keep your stack |
| Coding-agent-native | Via the framework | Native MCP, no framework buy-in |
| Code symbol graph | No | Yes |
| Bi-temporal truth + supersession view | Limited | First-class + Inspector diffs |
Competitor pricing and features change. Verify current figures against each vendor's live pricing page.
Free core, forever. Flat per-seat for the rest.
A hosted memory cloud meters you per memory and per retrieval. Engraphis is a flat license for software that runs on your box — the engine is free, and you pay only when memory becomes something a team needs to observe, prove, and govern. Never by the lookup.
Free
Best for a solo developer
- The complete recall engine — unlimited, on your machine
- MCP server, dashboard with Memory Inspector, CLI, library
- Yours under Apache-2.0, free forever
Pro
Best for proving what your agent remembers
- Everything in Community
- Analytics: growth, retention, and decay forecasts
- Compliance export — full bi-temporal JSON
- Priority support
Team
Best for a team sharing one memory
- Everything in Pro
- Multi-user Inspector — logins, roles, seats
- Shareable analytics HTML reports
- SSO/RBAC, audit exports, SLA
License keys are verified offline (Ed25519-signed), no phone-home. Pro and Team start with a 3-day free trial. Community is free forever.
Give your agents a memory.
> free and open-source at the core, runs entirely on your machine