The Ecosystem · TWO BINARIES, ONE GRAPH

Two Tools.One Experience.

The CLI is an agent that executes. Mind is a daemon that learns, and over MCP it becomes an experience server for Claude Code and any MCP client. They’re separate binaries on purpose: run either one without the other, and Mind will happily serve a client that isn’t Anvaya at all.

The Split

Why Two Binaries.

Anvaya Harness

A single static Rust binary, no runtime dependencies · 25.6 MB median RSS measured across 238 runs, 17x lighter than Claude Code · ~5 MB idle, ~12 ms to first TUI frame · 8 TUI screens · 21 native tools · 3 provider transports · headless mode.

Learn more →

◆

Anvaya Mind

An experience store, not a document · SQLite graph + Int8 HNSW index · 6 node types with per-type decay · 9 background workers · 14 MCP tools · captures debugging sessions, domain knowledge and project history.

Learn more →

Three Ways In, Same Graph.

Mind exposes the same memory through three transports. Which one you get depends on who is asking, and the cost difference between them is the difference between a function call and a protocol handshake.

<0.5msDirectMind runs in-process inside the CLI. The fastest path, used when anv owns the session and no other client needs the graph.

~2msIPCJSON-RPC over a Unix domain socket to the daemon. Multiple clients share one graph, with the daemon as the single writer.

~10msMCPThe daemon behind an MCP server. This is how Claude Code and other MCP clients reach the same memory.

Keep Your Agent. Add The Memory.

If you already work in Claude Code or another MCP client, you don’t have to adopt the CLI to get the memory. Point the client at Anvaya’s MCP server and fourteen memory tools appear inside the agent you already trust.

mind_queryPacked context for a query, under a token budget

mind_searchKeyword or semantic node search

mind_record_changeRecord a code change event

mind_get_nodeFetch one node by ID

mind_feedbackReport which injected nodes a turn used

mind_statsGraph size and health

mind_start_sessionOpen a session

mind_end_sessionClose it with an outcome

mind_get_threadsTemporal threads across sessions

mind_deltaWhat changed since a session

mind_proactiveUnprompted candidates under an EVOI gate

mind_intentThe session's inferred intent

mind_retro_promoteRecover suppressed knowledge

mind_federateOpt-in export/import across projects

Setup is a five-line .mcp.json in your project root, see Get Started →

Run Agents That Fit On Your Laptop.

25.6 MB median RSS. 25 agents ran in parallel on a Core 2 Duo with 4 GB RAM. Hundreds on your machine. Zero cloud required on the Ollama path.

Get Started →Read the FAQ

Requires Rust/cargo to build from source. Linux and macOS today, Windows not yet supported. Pre-1.0, public beta. Pricing TBD.