Now Available — v0.3.1

Your AI re-reads
everything. Every time.

AI keeps re-reading your entire codebase on every query. Kodara makes it remember — giving your AI surgical context instead of a firehose.

94% fewer tokens · Same answers · Any AI tool

The quality isn't in the model. It's in the context. · Senior engineer knowledge. Free model. One prompt.

See how it works

Kodara / Context Engine v0.3.1

Context
Engine

Query

kodara ask "How does auth work?"

Context Result — 3 modules / 1,840 tokens

auth/middleware.py — AuthMiddleware, verify_token()

auth/jwt_service.py — JWTService, decode()

models/user.py — User, UserSession

94.2% Reduction vs Full Read

00 / Problem

The Problem with
AI Coding Today

Every query re-reads everything

41,160 tokens per query

AI reads all 44 files → $0.62 per query

No memory between sessions — starts fresh every time

Breaks existing architecture patterns without knowing them

Token costs multiply fast at team scale

With Kodara

2,400 tokens per query

Reads 2 relevant files → $0.04 per query

Structured project memory that persists across sessions

Architecture-aware context — AI understands your system

94% cost reduction

01 / Features

What Kodara does

01 / Context Engine

Surgical Context

Finds the 2–8 most relevant files for any query using IDF-weighted semantic search + dependency graph expansion. No hallucinations from irrelevant context.

kodara ask "How does payment work?"
→ Found 6 modules / 1,840 tokens
→ billing/processor.py   ★★★★★
→ billing/stripe.py      ★★★★

02 / Intelligence

Impact Analysis

Know what breaks before you change anything. BFS traversal through the reverse dependency graph surfaces every affected module, ranked by risk.

kodara impact auth/middleware.py
Risk: HIGH
Direct affected:    2 modules
Indirect affected:  8 modules

03 / Memory

Accumulated Memory

Project memory that grows over time. Git history, developer annotations (the WHY layer), and architecture snapshots. Answers questions no file can.

kodara note add auth/jwt.py \
  "Using RS256 not HS256 — audit req"
kodara snapshot → saved 2026-03-12

04 / Onboarding

Instant Onboarding

New developer? Kodara generates a reading guide based on dependency layers and git activity. From zero to productive in hours, not days.

kodara onboard
01 → main.py           (entry point)
02 → api/server.py     (core)
03 → auth/middleware.py (foundation)

02 / Numbers

Real numbers.
Real project.

Tested on 44-file Python codebase · GPT-4o $2.50/1M input tokens (2026 pricing)

Without Kodara — 5 queries

$0.51

205,800 tokens consumed

AI reads all 44 files every time

41,160 tokens per query

$0.10 per query

No architectural awareness

With Kodara — 5 queries

$0.02

9,200 tokens consumed

Reads 2–3 most relevant files

1,840 tokens per query avg

$0.005 per query

94% reduction

Total saved on 5 queries $0.49

94% reduction

Monthly savings per developer · GPT-4o API

Light — 20 queries/day

$59

per month · $708/year

Medium — 50 queries/day

$147

per month · $1,764/year

Heavy — 100 queries/day

$295

per month · $3,540/year

Kodara Pro costs $19/month. Medium API user saves $147/month on GPT-4o token costs alone. Your return on investment: 7.7× every month.

7.7× ROI

Claude Opus · $5/1M

15.5×

ROI · $295/month saved · medium

Cursor Max Mode · $3/1M

10×

ROI · $190/month saved · medium

Team of 5 · GPT-4o

$735

saved/month · medium usage

03 / CLI

Every command
you need

Core

kodara init

Scan project, build memory

kodara ask "..."

Query project memory

kodara context

Generate ARCHITECTURE.md

kodara status

Index health & stats

Intelligence

kodara impact <file>

What breaks if I change this?

kodara onboard

Reading guide for new devs

kodara review

Impact of uncommitted changes

kodara diff <base>

Changes vs branch or commit

Memory

kodara history <file>

Git evolution & authors

kodara note add ...

Annotate files with WHY

kodara snapshot

Save memory state

kodara stats

Token savings report

04 / MCP

AI understands
your project. Always.

Without Kodara MCP

User:  How does the payment flow work?

Claude: [reading 83 files... 41,160 tokens]

        I'll analyze your codebase to understand...
        [repeating the same analysis every time]

With Kodara MCP

User:  How does the payment flow work?

Claude: [calling kodara_ask("payment flow")]
        → 4 modules / 1,840 tokens

        billing/processor.py handles Stripe charges.
        billing/webhook.py processes async confirmations...

One command.
Works everywhere.

Auto-configures Claude Desktop, Cursor, and VS Code simultaneously. No JSON editing. Restart your AI tool and Kodara is active.

kodara mcp --setup

Claude Desktop

Cursor

VS Code

Windsurf

What your AI gains

kodara_ask       — relevant modules for any query
kodara_impact    — what breaks if file X changes
kodara_context   — full architecture overview
kodara_search    — keyword search across all modules
kodara_status    — index freshness check

kodara_ask

Returns 2–8 most relevant modules for any natural language query. AI calls this automatically instead of reading the whole codebase.

kodara_impact

Shows which modules will break if a specific file changes. Risk-scored and depth-ranked.

kodara_context

Returns the full architecture map — hub modules, dependency layers, and Mermaid diagram.

kodara_search

Keyword search across all indexed modules and their summaries. Finds implementations by concept.

kodara_status

Index health check. AI uses this to verify the memory is fresh before querying.

05 / Extension

All commands.
Right in your editor.

Every Kodara command available as a sidebar button — no terminal switching. Free, Pro, Team, and Enterprise tiers visible at a glance. Pro and above show an upgrade prompt until you activate.

VS Code

Cursor

Windsurf

Trae

Install from Marketplace

Kodara — Actions

$(database) Init Project

$(search) Ask

$(info) Show Status

$(book) Architecture

$(history) File History

$(save) Save Snapshot

$(graph) Token Savings

PRO — $19/month

$(zap) Impact AnalysisPro

$(question) Why This FilePro

$(play-circle) Simulate ChangePro

$(flame) HeatmapPro

TEAM — $25/user

$(person) Onboard GuideTeam

$(warning) Bus FactorTeam

$(shield) Check ArchitectureTeam

06 / Pricing

Simple pricing.
Immediate ROI.

Free

forever

AI understands your project in seconds

Annotate files with context for AI

Point-in-time memory snapshots

Up to 200 files · unlimited queries

See your exact token & money savings

kodara stats — free, always

Today queries: 12 saved: $7.31
All time tokens: 2.3M saved: $35

Start Free

Pro

Your AI re-readseverything. Every time.

The Problem withAI Coding Today

What Kodara does

Real numbers.Real project.

Every commandyou need

AI understandsyour project. Always.

All commands.Right in your editor.

Simple pricing.Immediate ROI.

Start in30 seconds.