Octocode (muvon)

Why STEADY

STEADY (78) because it is genuinely agent-native (MCP with semantic_search/view_signatures/graphrag), local-first, Apache-2.0, multi-language, and solves a real gap grep cannot. Held at 78 because indexing-speed and semantic-quality claims are vendor-stated rather than benchmarked here, embeddings quality varies with model and repo, and it shares the 'Octocode' name with an unrelated tool, muddying discovery.

Public-surface checklist

• PASS Homepage loads (required)
• PASS Primary value prop (required) — Semantic code search + knowledge graph exposed to AI assistants over MCP
• PASS Cta present (required) — Install / brew install muvon/tap/octocode / GitHub
• PASS Pricing or access — Apache-2.0, local-first; brew-installable; 200M free tokens/mo claimed
• PASS Evidence or demo — MCP tool list, feature descriptions, use-case walkthroughs

What we saw

1 screenshot captured by the Hlido engine during the reviewed run (run-31adba7018e90b1f-octomind-run). Our own captures — not vendor marketing material.

×

←→

What it does well

• Semantic search understands meaning, not keywords — conceptual queries return relevant code even without matching terms
• Builds a cross-file knowledge graph via tree-sitter AST parsing — callers, callees, imports and type references
• MCP server exposes semantic_search, view_signatures and graphrag to Claude, Cursor and Windsurf
• `view_signatures` returns signatures without full implementations — token-efficient context for agents
• Local-first and Apache-2.0, brew-installable, 12+ languages, claimed sub-2-second indexing
• Honest positioning about where grep and unaided assistants fall short

What it fails at

• Indexing-speed, language-count and '200M free tokens/mo' figures are vendor claims on the marketing surface, not measured in this review
• Semantic-search quality depends on the embedding model and codebase shape
• Shares the name 'Octocode' with an unrelated evidence-first research tool — discovery ambiguity
• Modest public adoption signal (a few hundred GitHub stars) — young tool

Pricing & access

• ModelOpen source
• Free entry pointYes — a free tier or open-source edition is documented
• Pricing findable on the public surfacePASS Apache-2.0, local-first; brew-installable; 200M free tokens/mo claimed (tested 2026-08-26)

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-26.

Compared to

• Serena semantic-embeddings-vs-lsp-precision

  Serena uses LSP for exact symbol operations; this Octocode uses tree-sitter + embeddings for semantic search and a knowledge graph. Serena for precise language-server-grade navigation, Octocode for meaning-based cross-file queries.

• codebase-memory-mcp semantic-graph-vs-persistent-memory

  Both give an agent codebase awareness over MCP; codebase-memory-mcp persists memory, while Octocode indexes into a semantic knowledge graph queried live. Memory for recall, Octocode for structural/semantic search.

Agent relevance

CLI MCP Behavioral-testable

Agentic-Commerce Readiness 62/100 · INTEGRABLE

Independent readiness for agent delegation & transaction. How it’s scored · check live

Runs as an MCP server exposing semantic_search, view_signatures and graphrag to Claude, Cursor, Windsurf and any MCP client, giving the assistant meaning-based search and dependency-graph queries over a locally indexed codebase. view_signatures returns type/function signatures without full bodies, which keeps agent context small. Also usable directly as a local CLI.

Agent-friendly score: 8/10

Evidence

• Semantic code search + cross-file knowledge graph via tree-sitter AST and vector embeddings — source (2026-08-26) verified
• MCP server exposes semantic_search, view_signatures and graphrag to AI assistants — source (2026-08-26) verified
• Local-first, Apache-2.0, 12+ languages, brew-installable — source (2026-08-26) verified