Vs · OpenCode

Anvaya vs OpenCode.

The open-source multi-provider TUI with 75+ models, against the Rust agent that remembers. Same spirit, different bet.

Direct answer

Anvaya vs OpenCode: both are terminal agents that refuse model lock-in: OpenCode routes 75+ providers through a TypeScript TUI (plus desktop and IDE extensions), Anvaya runs a local-capable loop with Anthropic, OpenAI-compatible gateways and Ollama from a single static Rust binary. The split is memory: OpenCode keeps checkpoint summaries, so session 50 costs what session 1 cost; Anvaya keeps a compounding knowledge graph (typed nodes, decay, drift detection, outcome calibration) targeting 2–4K packed tokens per turn. OpenCode wins on provider breadth, surface coverage, and packaged maturity today; Anvaya wins on local footprint (~5 MB idle, no runtime), code-native retrieval (repomap PageRank + pooled LSP), and context that improves with use. Anvaya is pre-1.0: OpenCode is the mature open default; Anvaya is the memory bet.

The Table

Side by Side.

DimensionOpenCodeAnvayaWinner
Providers75+ via Models.devOllama + 3 + customOpenCode on breadth
Memory modelCheckpoint summaryTyped graph, calibratedAnvaya on compounding
SurfacesTUI + desktop + IDETUI + headless CIOpenCode on coverage
FootprintTS runtime + extensionsStatic binary (14–19 MB), ~5MB idleAnvaya on weight
Harness RSS (Phase 1)582.3 MB median25.6 MB medianAnvaya (measured, 238 runs)
LSPNativeNative pooled sessionsTie
LicenseMIT open sourceNot stated during betaOpenCode on clarity
Local modelsVia Ollama endpointsLocal endpoint, offline-capableAnvaya on local

Questions

Asked About the Two.

Q

How is Anvaya different from OpenCode?

OpenCode is the most flexible open-source harness: 75+ providers, subagents, plan mode, LSP, bring any model. Anvaya shares that philosophy (multi-provider, subagents, plan mode, native LSP) and adds what OpenCode lacks: a persistent memory graph that compounds across sessions, repomap PageRank retrieval, and a single static Rust binary (14–19 MB) with ~5 MB idle.

Q

Can I use OpenCode's models with Anvaya's memory?

Conceptually yes: Anvaya is multi-provider (Ollama, OpenAI-compatible endpoints, Anthropic), so the same Claude/GPT keys work. During beta the practical bridge is MCP, serve Anvaya's memory to any harness while the native multi-provider loop hardens.

Q

Which is lighter?

Anvaya: one static Rust binary (14–19 MB), ~5 MB idle, no runtime. OpenCode ships a TypeScript TUI plus desktop and IDE extensions, broader surface, heavier footprint. Pick OpenCode for surface coverage, Anvaya for minimal local footprint plus memory.

Q

How open are Anvaya's formats?

Anvaya builds from source and the memory files are open formats (SQLite, flat index, NDJSON log) with visible tool and context decisions; licensing is not stated during beta. OpenCode is MIT with a broad packaged distribution.

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.