Local-first · Free

Your projects. Your memory. Your AI. 

Git Lora is a local-first workspace that gives AI long-term project memory, knowledge graphs, agents, companions and autonomous workflows.

Personal is free. Your data stays under your control.

AI forgets. Projects fragment.

Today's AI tools are powerful but amnesiac. Each session starts from zero, context is expensive, and your work is scattered across clouds.

01

AI forgets

Conversations reset. Long-term project context never accumulates.

02

Context is expensive

Re-sending everything every time costs tokens, time and money.

03

Information fragments

Notes, code and ideas live in disconnected tools.

04

Projects disconnect

Story, structure and code never share one source of truth.

05

Cloud-only memory

Private work lives on someone else's server by default.

One workspace. Everything connected.

Files, Markdown, code, entities, a knowledge graph, memory and AI — in a single local-first project.

Files & Markdown

Plain-text first. Yours, readable, portable.

Code

First-class code files with AI assistance.

Entities

Characters, locations, companies, events, objects.

Graph

Visualize relationships and project structure.

Memory

Persistent, scoped, inspectable project memory.

AI chat

A companion that knows your whole project.

Seven memory capabilities

Git Lora gives AI a real memory subsystem — not a growing prompt, but a managed layer.

01

Automatic Memory Router

Routes facts to the right scope automatically.

02

Automatic Memory Extraction

Pulls durable facts from conversations as you work.

03

Temporal Memory

Knows what was true then and what is true now.

04

Contradiction Detection

Flags when new information conflicts with stored facts.

05

Memory History

Every memory is versioned and recoverable.

06

Memory Scopes

Project, companion and global scopes stay isolated.

07

Memory Inspector

Read, edit and debug exactly what the AI remembers.

gitlora — memory

A companion that proposes — never overwrites

Chat with characters or coding experts using your project as context. Conversations stay separate from project state. Every AI change becomes a proposal you control.

For writers

Choose a character, chat in character, keep companion memory separate from story state.

For developers

Choose an expert for debugging, architecture advice and refactors.

companion · proposal
// scene — chapter 7
Kira stepped into the archive, the air thick with
old paper and quiet electricity.
AI proposesShe paused at the third shelf, fingers tracing a name she had tried to forget.

An edit, a new scene, a new entity, a relationship, a refactor, a doc update — you decide.

A graph that mirrors your project

Entities, relations, branches, story development and project architecture — connected and navigable, node to node.

gitlora — graph
ProjectMemoryCodeFilesEntitiesCompanionAgentEventsThoughts
Node

Project

Connected · 8
  • Memory
  • Code
  • Files
  • Entities
  • Companion
  • Agent
  • Events
  • Thoughts

Drag a node to move it — the rest of the graph reacts. Hover to isolate a node and its connections.

Agent OS

Orchestrate multiple agents with explicit permissions, memory scopes, an event bus, scheduler, checkpoints and retry/resume.

agent · 01

ChatZ / GLM

Primary agent for heavy, high-quality work.

agent · 02

Local Llama

Auxiliary local agent for privacy-sensitive and lightweight tasks.

agent · 03

QC Agent

Quality-control agent that reviews before you commit.

agent · 04

Permissions

Per-agent tool permissions. Nothing runs unchecked.

agent · 05

Memory scopes

Each agent sees only the memory it should.

agent · 06

Workflows

Event bus, scheduler, checkpoints and retry/resume.

Autonomous Engine

Long-running workflows that plan, execute, checkpoint, recover, run QC and continue — then hand off to the Finalizer.

  1. 1
    Plan

    Decompose the goal into steps

  2. 2
    Execute

    Run each step against your tools

  3. 3
    Checkpoint

    Persist progress to memory

  4. 4
    Recover

    Retry or reroute on failure

  5. 5
    QC

    Verify outputs against the goal

  6. 6
    Continue

    Advance to the next step

  7. 7
    Finalizer

    Assemble the finished artifact

gitlora — engine.log
›PlanDecompose the goal into steps
›
runningstep 1/7

For local autonomous processing, keep the computer running.

Finalizer

Turn project material into a finished artifact. Git Lora proposes a structure for your approval, then generates the output.

Book
Screenplay
Report
World bible
Documentation
Code package
Structured export
Custom format

Local-first by design

Project content stays on your machine by default. No mandatory cloud memory. Optional, user-controlled sync. Your AI provider, your choice.

  • Project data is local by default
  • No mandatory cloud memory
  • Optional user-controlled sync server
  • Configurable AI provider
  • Works offline when gitlora.uz is unavailable

Built for developers

First-class support across the modern stack — and extensible to more.

PythonJavaScriptTypeScriptHTML/CSSReactNext.jsVueJavaKotlinC#C/C++RustGoPHPRubySQLBash/PowerShellYAML/JSON/TOMLMarkdownand extensible language support

Personal vs Business

Git Lora Personal

Free
  • Download and use
  • Local-first
  • Community support
  • Full project model
Download Git Lora

Git Lora Business

Request access
  • Organization onboarding
  • Commercial terms
  • Dedicated support
  • Deployment & integration options
Request Git Lora Business

Verified traction

We show real milestones — not fabricated numbers. Metrics appear here as they are verified.

M01

Public release

Git Lora Desktop shipping for Windows, macOS and Linux.

M02

Memory subsystem

Seven-capability memory layer operational.

M03

Agent OS

ChatZ/GLM, local Llama and QC agents orchestrated.

M04

Open roadmap

Ideas, changelog and traction tracked in public.

Downloads, active installations and country counts will appear once verifiably collected with consent.

View all →

Build projects that remember.

Start free. Stay local. Keep control.