Skip to main content
CodeAlive

RepoQA benchmark: -45% tokens at a higher score

The context engine for AI agents on large codebases

Semantic search plus a live code graph of calls and dependencies, served over MCP and API to Cursor, Claude Code, Codex, and any agent you run. Precise, cited context with 45% fewer tokens, at ~25× lower model cost.

Playground requires no signup - Index your first repo in minutes

−45%
fewer tokens
Same 20 RepoQA tasks, same model — semantic_search off vs. on.

Works with the tools you already run

GitHubGitLabBitbucketAzure DevOpsCursorClaude CodeCodexWindsurfZedVS CodeJetBrainsContinueCline
GitHubGitLabBitbucketAzure DevOpsCursorClaude CodeCodexWindsurfZedVS CodeJetBrainsContinueCline
GitHubGitLabBitbucketAzure DevOpsCursorClaude CodeCodexWindsurfZedVS CodeJetBrainsContinueCline
GitHubGitLabBitbucketAzure DevOpsCursorClaude CodeCodexWindsurfZedVS CodeJetBrainsContinueCline

Index. Retrieve. Ground.

Your repos become a queryable map of the system — and every agent answer comes back grounded in it.

  • 01

    Index

    Connect your repos. CodeAlive builds a code graph — files, symbols, calls, dependencies, tests — and keeps it fresh on every push.

  • 02

    Retrieve

    Every query runs hybrid retrieval: semantic search + lexical match + graph traversal. Results carry relationships, not just similarity.

  • 03

    Ground

    Agents call CodeAlive over MCP or API and get precise, cited context: the right files, the call chain, and the tests that cover it.

The code graph behind every answer — entities and the relationships between them.

Live Indexing

One context engine. Three ways to use it.

Every surface queries the same code graph — pick the one that fits your workflow.

Ground any coding agent

Add the CodeAlive MCP server to Cursor, Claude Code, or Codex — or call the REST API from your own agents and CI. One endpoint, the whole codebase.

Explore MCP & API
.mcp.json
{
  "mcpServers": {
    "codealive": {
      "url": "https://mcp.codealive.io/api/"
    }
  }
}
// tools: semantic_search · grep_search · chat
One endpoint, the whole codebase

Ask multi-step questions across repos

Deep Research plans, traverses the graph across services, and returns a cited report — architecture, data flows, and the impact of a change.

Explore Deep Research
  1. Parse question

    Decompose 'How does our payment flow work?' into retrievable sub-queries.

  2. Retrieve context

    Pull PaymentService, OrderRouter, and webhook handlers from the knowledge graph.

  3. Multi-hop reasoning

    Trace the chain from API entry to service to external vendor to confirmation.

  4. Cite & answer

    Grounded answer with file:line references back to your code.

Deep research, one step at a time

Review PRs with full-system context

The Code Review Agent runs on pull requests and catches changes that look safe in the diff but break flows two services away. Posts straight back to your Git provider.

Explore Code Review
services/checkout.ts+5 / -1
@@ -42,7 +42,12 @@ services/checkout.ts
  async function checkout(order) {
-   const total = order.items.reduce((s, i) => s + i.price * i.qty, 0)
+   const total = computeTotal(order.items)
+   if (!Number.isFinite(total)) {
+     throw new Error('invalid order total')
+   }
    return paymentService.charge(order.customerId, total)
  }

CodeAlive · review

This logic already exists in PaymentService.computeTotal(). Extract the helper instead of duplicating it.

AI review comment on a real diff

45% fewer tokens. Higher answer quality. Same benchmark.

Measured on RepoQA — not on a slide. Full methodology, prompts, and raw runs are public.

−45%

fewer tokens to answer the full RepoQA suite — from 3.99M down to 2.17M

+6.5 pts

higher answer quality, on roughly half the context

~25×

lower model cost vs. frontier agents on identical tasks

Read the full methodology

RepoQA: CodeAlive + Qwen3.6 deep vs Claude Opus 4.8 max / Codex GPT-5.5 high. Real results vary by repository and task mix.

Fits your stack. Stays inside your perimeter.

Connects to

Indexes any text-based source — TypeScript, Python, Go, Java, C#, Rust, and more.

Git
GitHubGitLabBitbucketAzure DevOpsGitea
Agents
CursorClaude CodeCodexWindsurfZedContinue
IDEs
VS CodeJetBrains
CI/CD
Any pipeline — review agent posts back to your Git provider

Runs inside your perimeter

Cloud by default. Fully self-hosted when your code can't leave the building.

  • Your code stays put — self-hosted deployments never send code outside your network.

  • Bring your own LLM — Qwen, Llama, gpt-oss, or any OpenAI-compatible endpoint.

  • Encrypted at rest — AES-256-GCM with per-organization keys; we never train on your code.

  • Docker Compose · Kubernetes · Helm — deploys like the rest of your infra.

What our customers are saying

We have a pretty complex codebase that tools like Cursor and DeepWiki couldn't quite figure out on their own. I decided to index our project with CodeAlive and was pleasantly surprised, as its responses demonstrated a deeper understanding of our project's logic.
Alexander KolotovBlockscout
The CodeAlive's "Deep" chat really creates nice results. From a developer's perspective - joining a new project and getting their first tasks (e.g., add a new DDD Entity in a complex project) - the AI produces an accurate how-to. I like this!!!
Hauke FeddersenCTO @ grasbyte GmbH
During our work with CodeAlive, we've found that it's beneficial for our QA by helping with writing test cases and describing bugs. It also accelerates project onboarding. A very useful product!
Zhaksylyk UaliyevCTO @ Esqadra Technologies

Transparent usage-based pricing. Your monthly balance covers search, chat, deep analysis, and code review.

Search$0.02
Chat$0.15
Deep analysis$0.30
Code review$0.50

Try CodeAlive without signing up

Browse our hand-picked, pre-indexed open source repositories.

Open the Playground

Free

$0/ month

Free tier

Estimate with calculator
  • 1 user / 1 workspace
  • 25 MB total repo size
  • Public & private repos
  • 100 chat requests / month
  • MCP access

Get Started Free

For individual developers indexing a single small project.

Hobby

$15/ month

$15 monthly usage balance

Estimate with calculator
  • 1 developer
  • Up to 100 MB codebase
  • Semantic code search
  • Chat with your code
  • Set your own spending cap
  • ≈100 chats or ≈750 searches

Get Started

For a solo developer who wants semantic search and chat on a focused codebase.

Most Popular

Pro

$50/ month

$50 monthly usage balance

Estimate with calculator
  • 1 - 3 developers
  • Up to 100 MB codebase
  • Deep analysis
  • AI Code Review
  • Spending cap & usage alerts
  • ≈330 chats or ≈2,500 searches
  • Everything in Hobby

Get Started

For developers and small teams that need deep analysis, AI code review, and more usage.

Team

From $100/ month

$100+ monthly usage balance

Estimate with calculator
  • 5 - 15 developers
  • 100 - 500 MB codebase
  • Team spending controls
  • Usage analytics dashboard
  • Priority support
  • ≈5,000+ searches / month
  • Everything in Pro

Contact Sales

For engineering teams with larger codebases, team controls, analytics, and priority support.

Frequently asked questions

Still have questions? Email us at support@codealive.io

Give your agents the whole codebase

Index your first repo in minutes — or try the Playground without signing up.