Harmony MCP • High-Performance Agentic Memory

Cure Codebase Amnesia

AI agents waste time rediscovering your codebase. Harmony gives them high-performance agentic memory so they spend more time coding.

$npx @codeconductor/harmony install
  • No account required
  • First repository free
  • Install in under 30 seconds
90%
Less searches
10×
Lower token costs
60ms
Faster memory retrieval
claude code · harmony mcplive
agentresolve context for this prompt…
↳ harmony.getBundle(prompt)
harmonycontext bundle ready~100ms
context bundle · 6 filesrelevance
auth.ts0.94
routes.py0.88
models.rs0.76
schema.sql0.69
index.ts0.61
94
CAB Score

Trusted by the world’s most innovative enterprises

UCSFTaxPlanIQConnect CRECPA PilotSeekFirst
The difference

Your AI shouldn’t have to rediscover your codebase.

Every new prompt starts another round of searching, grepping, and reading. Harmony remembers, so your agent spends its time solving problems instead of rediscovering your codebase.

Without Harmonyagent + filesystem
# resolving context…
list files  read  decide
read more  re-read  decide
grep  read  decide
↻ ~14 round-trips
several seconds · thousands of tokens
With Harmonyagent + harmony
# resolving context…
harmony.getBundle(prompt)
↳ context bundle ready

✓ 60ms · 0 extra trips
10× fewer tokens · 4× faster
The engine

Built to eliminate Codebase Amnesia

From indexing and retrieval to formatting and token budgeting, every part of Harmony is engineered to reduce rediscovery and maximize useful context.

Hyper-Converged Contextual Indexing

Harmony unifies semantic search, symbol resolution, call graphs, imports, and recent code changes into a single memory index. Multiple ranking passes ensure every context bundle is complete, relevant, and compact.

Proprietary · Multi-pass ranking

Token Budgeting

Harmony intelligently packs the highest-value context into every request, eliminating unnecessary retrieval and keeping token usage predictable. More tokens go toward solving problems, not rediscovering your codebase.

10× fewer tokens

Adaptive Context Expansion

When more context is needed, Harmony expands outward from the most relevant code: following callers, callees, symbols, and dependencies until the agent has exactly what it needs.

Dynamic context expansion

Built for LLMs, Not APIs

Most memory servers serialize context for machines. Harmony returns structured markdown designed for language models, improving comprehension while reducing wasted tokens.

Model-optimized format

Repository Isolation

Every repository is stored in its own isolated memory space with strict tenant separation. Harmony never mixes customer repositories, ensuring every agent only accesses the codebase it’s authorized to use.

One repository. One memory.
Coming soon

Model-Aware Context

Claude Code, Cursor, Windsurf, Gemini, and future agents all consume context differently. Harmony automatically formats every context bundle for the model that’s asking.

Optimized per model
How it works

Context that reasons, retrieves, and remembers.

Most MCP memory servers retrieve context. Harmony builds it. Hyper-Converged Contextual Indexing continuously understands your repository, expands only when needed, optimizes every request, and delivers context tailored for the coding agent using it.

1

Hyper-Converged Contextual Indexing

Harmony continuously builds a living understanding of your repository, connecting code, relationships, and recent changes into a single high-performance memory that’s always relevant, compact, and ready for every prompt.

Building repository memory...
✓ Repository analyzed
✓ Relationships connected
✓ Context optimized
24,831 items indexed
Proprietary · Multi-pass indexing
2

Adaptive Context Expansion

Harmony starts with the smallest relevant context, then expands only when needed. Instead of searching your entire repository, it follows the most relevant paths until the agent has exactly what it needs.

Resolving context...
auth
middleware
permissions
✓ Context complete
Dynamic context expansion
3

Intelligent Token Budgeting

Harmony fills every request with the highest-value context that fits within your model’s token budget. Less wasted retrieval. More useful reasoning. More tokens spent writing code.

Packing context...
31,872 / 32,000 tokens
✓ Budget optimized
~10× fewer tokens
4

Agent-Aware Context

Coming soon

Every coding agent thinks differently. Harmony automatically formats every context bundle for Claude Code, Cursor, Windsurf, Gemini, and future agents, improving accuracy while reducing unnecessary tokens.

Target agent
Claude Code
Bundle ready
✓ Markdown
Optimized per agent
Why Harmony

Most MCP memory retrieves. Harmony remembers.

Traditional MCP memory servers retrieve files. Harmony builds a living understanding of your repository, allowing coding agents to stop rediscovering code and start solving problems. Every prompt begins with the right context instead of another round of searching.

Traditional MCP MemoryHarmony
Retrieves relevant filesUnderstands the repository
One retrieval passMulti-stage context construction
Static contextAdaptive context expansion
Generic outputAgent-aware context bundles
Repeats discovery every sessionPersistent repository memory
Token-heavy retrievalToken-budgeted reasoning

Traditional memory servers help agents find information.

Harmony helps agents remember it.

The result is fewer searches, lower token costs, faster responses, and more accurate code generation.

The Codebase Amnesia Benchmark

The first benchmark built for AI coding agents.

Most benchmarks measure language models. CAB (Codebase Amnesia Benchmark) measures something different: how efficiently coding agents understand, retrieve, and reuse knowledge inside real software projects. Compare any coding agent before and after Harmony using the same repository, the same prompts, and the same tasks.

Rediscovery Rate
Measures how often an agent repeats searches for information it has already discovered.
Search Depth
Measures how many search and file exploration steps occur before meaningful work begins.
Knowledge Reuse
Measures how effectively an agent remembers and reuses previously discovered information.
Context Acquisition Cost
Measures the percentage of tokens spent rebuilding context instead of solving the task.
Time to First Meaningful Edit
Measures how quickly an agent starts writing code instead of searching the repository.
Codebase Amnesia Benchmark (CAB)
WorkflowCAB Score
Claude Code
61
Claude Code + Harmony
93
Cursor
64
Cursor + Harmony
95
Codex
58
Codex + Harmony
91

Same coding agent. Same repository. Same prompts. The only difference is Harmony.

Read the CAB Methodology
Customer stories

Trusted by teams already building with AI agents.

Harmony compressed what would have been months of context-setup work into a single afternoon. Our Claude Code sessions now start with full repo awareness from the first prompt. No warm-up, no repeated file reads. The token savings alone paid for the switch.

HM
Harsh Mody
CEO, CPAPilot

We built our entire AI coding workflow on MCP-connected agents. Harmony was the missing piece. It gave our agents the codebase memory they needed to stop re-discovering the same modules on every task. Landmark Expansion is a game-changer for our monorepo.

SM
Sanjeev Malhotra
Founder, 5plus

Token waste was quietly eating our Claude API budget. After installing Harmony, Cursor sessions are dramatically faster and we have seen a measurable drop in redundant API calls. The XML + Markdown format makes a real difference. Our agent actually understands the context it receives.

LH
Latif Hamlani
Founder, PartnerAssistant.ai
Quickstart

One command. Every coding agent.

Harmony speaks the Model Context Protocol (MCP), making installation nearly identical across Claude Code, Cursor, Windsurf, Codex, Gemini, and every MCP-compatible coding agent. Install in seconds. No account required. Your first repository is free.

# Install Harmony into your current repository
npx @codeconductor/harmony install
✓ Repository detected
✓ Building repository memory...
✓ Harmony ready
Codebase Amnesia cured.
  • No account required
  • First repository free
  • Ready in under 30 seconds
  • Works with every MCP-compatible coding agent
Pricing

You don’t need to choose a plan today.

Install Harmony with no account required and your first repository is completely free. Need another repository? Create a free account in one click. Harmony only asks you to upgrade when your projects grow.

Free

$0

For personal projects and evaluation.

  • Everything from anonymous install
  • 2 repositories
  • Free forever
  • Any MCP-compatible coding agent

Developer

$7/mo
$70/year

For developers building across multiple repositories.

  • 10 repositories
  • Commercial use
  • Priority memory serving
  • Faster indexing

Team

$29/user/mo
$290/user/year

Shared memory for engineering teams.

  • Unlimited shared repositories
  • Team repository memory
  • Team management
  • Priority support

Enterprise

Custom

For organizations with advanced security and deployment requirements.

  • Self-hosted deployment
  • SSO & SCIM
  • Private infrastructure
  • Dedicated support
  • Enterprise SLA

Already using Harmony?

Harmony automatically prompts you when it’s time to upgrade. Until then, just keep coding.

Available everywhere

One memory. Every coding agent.

Harmony works anywhere the Model Context Protocol does. Install it once, then use the same high-performance repository memory across Claude Code, Codex, VS Code, Cursor, Windsurf, Gemini, and every MCP-compatible coding agent.

Native Integrations

Native plugins for Claude Code, Codex, and VS Code deliver the fastest installation experience. One click and Harmony is ready to build memory for your repository.

Claude Code · Codex · VS Code

MCP Ecosystem

Harmony speaks the standard Model Context Protocol, making it compatible with Cursor, Windsurf, Gemini, ChatGPT, OpenAI and every MCP-compatible coding agent. Install once. Your agents share the same repository memory.

Smithery · Glama · PulseMCP
Coming soon

Deploy Anywhere

Run Harmony locally, in your own cloud, or self-hosted for enterprise deployments. Container support and cloud marketplaces make rollout simple for teams of every size.

Docker · AWS · Azure · DigitalOcean
Get started

One command later, your agent remembers.

Install Harmony in under 30 seconds. No account required. Your first repository is completely free. Add a second repository whenever you’re ready with a free Harmony account.

$npx @codeconductor/harmony install
  • No account required
  • First repository free
  • Ready in under 30 seconds
Common questions

Questions teams ask us.

Everything you need to know about Harmony before your first install.

Codebase Amnesia is when an AI coding agent repeatedly searches, reads, and rebuilds context it has already discovered. Those repeated discovery cycles waste tokens, increase latency, and reduce code quality. Harmony gives coding agents persistent memory so they spend less time rediscovering your repository and more time writing code.
Harmony MCP for Codebase Memory | CodeConductor