MCP-native code intelligence

The MCP layer
for your codebase.

ACECode indexes your code once and serves it back through the Model Context Protocol to Claude, Cursor, VS Code, or any MCP-compatible AI — with semantic search, code graphs, git lineage, and learned conventions wired in.

Start free Read the docs

Free tier · 1 project · 5K credits/mo · no card required

~/your-project — claude
$ ace init && ace index
✓ Indexed 847 files · 12,840 chunks · 3,210 symbols in 42s
$ ace search "how does the auth middleware work?"
🔍 Found 8 results in 280ms — reranked with grok-4.3
→ packages/core/api/src/middleware/auth.ts:46
→ packages/features/auth/src/server/services/auth.service.ts:19
# Same query, now from Claude Code via MCP:
claude The auth middleware lives at packages/core/api/src/middleware/auth.ts. It validates Bearer tokens against the api_keys table, then attaches the user to the Hono context as c.var.user. The session token version comes from session.service.ts:42…

Built on MCP

One index. Every MCP client.

ACE is the canonical MCP server for code intelligence. Add it to Claude, Cursor, Continue, or any client speaking the Model Context Protocol — no glue code, no proxies, no per-client integrations.

Claude Code Official
Cursor MCP
Continue MCP
VS Code (ACE ext) Native
Zed MCP
Cline MCP
Any MCP client +
Setup guide →

Built for AI-assisted dev work

Six things your AI gets right with ACE.

Semantic + keyword hybrid

Voyage voyage-4 embeddings blended with BM25 full-text via tsvector. Reranked with Grok 4.3 on Max/Business. ~280-400ms per query.

Memories that survive sessions

ACE detects patterns from your code and recalls them across conversations. Stop re-explaining your conventions to every new chat.

Code graph + impact analysis

Map dependencies before refactoring. See the blast radius of a change. Trace call chains from any entry point. Powered by tree-sitter + TS Compiler API.

Multilingual queries

Tiếng Việt, 中文, 日本語 — query in any language. Voyage voyage-4-large + Grok cross-lingual rerank. Validated on 8 query languages.

Hallucination check

Validate any symbol or import path against the live index. Catch AI suggestions that reference functions that don't exist before they reach your code.

Drift-aware lineage

ACE tracks how every symbol evolved through git. "Why does this look this way?" becomes a one-call MCP query — with the commit, the author, and the rationale.

Where ACE lives

One index. Three surfaces.

Index your code once with the CLI. Use it from wherever you work.

CLI

Free

ace search, ace index, ace watch. Scripts, hooks, terminal queries — everything an MCP tool can do has a CLI surface.

App dashboard

Free

Browser UI at app.acecode.dev. Project status, billing, API key management, and a search playground.

ADE

Pro+

ACE Development Environment — a local-web IDE with intelligence widgets baked in. VS Code-clone shell, all surfaces unified.

Credit-based pricing

Pay for AI calls. Embeddings included.

1 credit ≈ $0.0001 of LLM cost. Free tier covers most solo dev usage.

Free

$0 /mo

5K credits / mo

  • 1 project
  • CLI + MCP
  • Community routing
Popular

Pro

$19 /mo

50K credits / mo

  • 10 projects
  • ADE access
  • Model preference

Max

$49 /mo

250K credits / mo

  • 50 projects
  • Grok 4.3 default
  • Higher pool tier

Business

$199 /mo

1.5M credits / mo

  • Unlimited projects
  • Self-host
  • SSO + SLA
How credits work →

Index once. Query forever.

Free tier: 1 project, 5K credits/month, no card needed.

Create free account Browse the docs