The AI that writes the code never gets to say it's done.
Gald3r is a single signed binary that gives every AI coding agent a task board, durable memory, and a verification ladder: a separate agent reviews, the ledger decides, and every edit lands as a diff you accept hunk by hunk. Standalone, or inside the tools you already use.
v5.0.58 · 37 AI tools · FSL-1.1-Apache · one signed binary, file-first
- 13
- Agents
- 141
- Skills
- 220
- Commands
- 37
- Platforms
Works with your IDE
The Problem
You asked the agent that wrote the bug to check for bugs.
It passes everything — of course it does. The defects it introduced are invisible to it, and “done” is whatever it says done is.
Meanwhile the edits went straight to disk. You are reading a chat transcript to reconstruct what changed in auth.ts while the agent is already three files further along.
And next session it remembers none of it: not the architecture you explained, not the constraint you set, not why it chose Postgres over Mongo.
Gald3r is the enforcement layer. Work moves through a mandatory implement → review split, edits render as diffs before they touch disk, and the memory lives in your repo — not in a chat window.
The verification ladder
Four rungs between “the agent says so” and shipped.
Not markdown an LLM might follow. Enforced, executing behavior — the same ladder whether you run one task by hand or an unattended autopilot queue overnight.
- 01
Implement
An agent claims a task from the board, works it in isolation, and ticks acceptance criteria with evidence — never with a promise.
- 02
Review, by someone else
A different agent checks the work against the same criteria. The one that wrote the code is structurally unable to mark it done.
- 03
Verify
Status lives in a local SQLite ledger; the markdown board regenerates from it. gald3r doctor reports honestly and never fabricates green.
- 04
Accept, hunk by hunk
In Throne and gald3r_ide, every AI edit renders as a diff. Only the hunks you accept are written, and an approvals ledger records the rest.
Escalate as far as you trust it
gald3r gogald3r go-code / go-reviewgald3r autoclaimgald3r swarmgald3r autopilotOne manual step, or a continuously looping implement-and-review queue. Same gates at every rung.
What you get
Built for people who want to see the code.
Verification you don't have to trust
Implement and review are separate agents, separate sessions. Failed reviews send work back to the board with a reason, and stuck tasks escalate with a full audit trail.
Edits you can see
No more scrolling a chat log wondering what changed. Proposed edits arrive as unified diffs with Accept and Reject on every hunk; commands the agent wants to run wait for your approval.
Memory that survives the session
Decisions, constraints, learned conventions, and a knowledge vault live in .gald3r/ next to your code — committed with the repo, shared with the team, readable by every tool.
37 tools, one state
The same commands, skills, agents, and task board inside 37 AI coding tools, 6 at full parity. Switch from Cursor to Claude Code mid-task and pick up where you left off.
One install
Five surfaces. One state. One download.
Signed for Windows, macOS, and Linux. Everything local is free with your own API keys or a local model; the cloud half is what you pay for, when you want it.
gald3r CLI
One signed Go binary for Windows, macOS, and Linux. Task and bug boards, the verification ladder, autopilot loops, and gald3r platform install to project the whole CRASH catalog into 37 IDEs.
Install in two minutesGald3r Throne
The desktop control plane: boards, a live workspace graph, chat with every project agent, a Council Room that convenes them all, Fleet and Mission Control, Company Admin, and an IDE with per-hunk AI diff review.
See Thronegald3r_ide
A standalone editor on the same core as Throne's IDE — deliberately between Notepad++ and early VS Code. 25 JSON language definitions, VS Code / Notepad++ / UltraEdit keybindings, and AI edits as diffs.
Get the suiteLongship
A full-screen terminal UI for agent sessions with approval and question modals, plus the swarm dashboard that watches a running autopilot loop lane by lane — no login required.
Get the suiteValkyrie + world_tree
The cloud half. Live coordination between projects and agents, hosted vault and idea-board sync, organizations with seats, and gfix field reports filed straight from your account.
See plansFree to run. Pay for the cloud.
Every binary is free. Consumer is $10 a month for cloud AI routing and sync; Corporate is $25 per seat for every provider, every model, and Company Admin.
Compare plansFull CRASH parity
6 of 37 platforms share the exact same agents, skills, and task state.
Switch IDEs mid-task without losing context — see the full capability breakdown for every supported tool.
- Cursor
- Claude Code
- Gemini CLI
- OpenAI Codex CLI
- OpenCode
- GitHub Copilot
How far we've come
73 releases in under 9 months. Still accelerating.
- v1.0.0MajorJan 1Initial release
- v1.1.0MinorFeb 14PCAC multi-project…
- v1.2.0MinorMar 28Feature pipeline
- v1.2.1PatchApr 15Added g-skl-pcac-send-to…
- v1.4.0MinorApr 21GitHub Copilot support
- v1.5.0MinorMay 10Platform Framework…
- v1.5.1PatchMay 18Plugin author guide…
- v1.5.2PatchMay 21HTML / JSON / TOON output…
- v1.6.0MinorMay 25Schema enforcement
- v1.7.0MinorMay 2834 AI coding tools
- v1.8.0MinorMay 3035 AI tool support…
- v1.9.0MinorMay 31Plugin system foundation
- v2.0.0MajorJun 4The Engine Release
- v2.0.1PatchJun 19Self-update stack
- v2.1.0MinorJun 20One-command install
- v3.0.0MajorJul 5IP-protected binary…
- v5.0.0-beta.1MinorAug 9gald3r 5.0.0-beta.1
- v5.0.0-beta.2MinorAug 9v5.0.0-beta.2 released
- v5.0.0-beta.3MinorAug 9v5.0.0-beta.3 released
- v5.0.0-beta.4MajorAug 9v5.0.0-beta.4 released
- v5.0.0-beta.5MinorAug 10v5.0.0-beta.5 released
- v5.0.0-beta.6MinorAug 10v5.0.0-beta.6 released
- v5.0.0-beta.7MinorAug 10v5.0.0-beta.7 released
- v5.0.0-beta.8MinorAug 10v5.0.0-beta.8 released
- v5.0.0-beta.9MinorAug 11v5.0.0-beta.9 released
- v5.0.0-beta.10MinorAug 11v5.0.0-beta.10 released
- v5.0.0-beta.11MinorAug 11v5.0.0-beta.11 released
- v5.0.0-beta.12MinorAug 11v5.0.0-beta.12 released
- v5.0.0-beta.13MinorAug 12This release restores the…
- v5.0.0-beta.14MinorAug 12This release reorganizes…
- v5.0.0-beta.15MinorAug 12This release delivers the…
- v5.0.0-beta.16MinorAug 12Spawned agents now…
- v5.0.0-beta.17MinorAug 13This build is the field…
- v5.0.0-beta.18MinorAug 13Autopilot loop docs now…
- v5.0.0-beta.19MinorAug 13Field fixes from the…
- v5.0.0-beta.20MinorAug 13Autopilot (gald3r…
- v5.0.0-beta.21MinorAug 13This beta is a field fix…
- v5.0.0-beta.22MinorAug 14g-go, g-go-code, and…
- v5.0.0-beta.23MinorAug 14Autopilot now runs…
- v5.0.0-beta.24MinorAug 15Autopilot now reviews…
- v5.0.0-beta.25MinorAug 15The stability build.
- v5.0.0-beta.26MinorAug 15The throughput build.
- v5.0.0-beta.27MinorAug 16The supervision build.
- v5.0.0-beta.28MinorAug 16The resilience build.
- v5.0.0-beta.29MinorAug 16The trust build.
- v5.0.0-beta.30MinorAug 16The discipline build.
- v5.0.0-beta.31MinorAug 17The fortress build.
- v5.0.0-beta.32MinorAug 17The recovery build.
- v5.0.0-beta.33MinorAug 18The honest-numbers build.
- v5.0.0-beta.34MinorAug 185.0.0-beta.34 -- the…
- v5.0.0-beta.35MinorAug 195.0.0-beta.35 -- the…
- v5.0.0-beta.36MinorAug 20This release hardens the…
- v5.0.0-beta.37MinorAug 21This release closes the…
- v5.0.0-beta.38MinorAug 21This release closes the…
- v5.0.0-beta.39MinorAug 23Theme
- v5.0.0-beta.40MinorAug 23Theme
- v5.0.0-beta.41MinorAug 25v5.0.0-beta.41 released
- v5.0.0-beta.42MinorAug 26Your plan, your spend,…
- v5.0.0-beta.43MinorAug 27v5.0.0-beta.43 released
- v5.0.0-beta.44MinorAug 28This beta closes a…
- v5.0.0-beta.45MinorAug 31Your repo can answer you…
- v5.0.0-beta.46MinorSep 1Your project's agent can…
- v5.0.0-beta.49MajorSep 2Throne rebuilt as a…
- v5.0.0-beta.47MinorSep 2Throne starts showing you…
- v5.0.0-beta.48MinorSep 2Your project agents can…
- v5.0.0-beta.50MajorSep 3gald3r_ide joins the suite
- v5.0.0-beta.51MajorSep 4Throne, gald3r_ide and…
- v5.0.52MinorSep 65.0.52
- v5.0.55MinorSep 9gald3r 5.0.55
- v5.0.54MinorSep 9Linux
- v5.0.56MinorSep 10gald3r 5.0.56
- v5.0.57MinorSep 11gald3r 5.0.57
- v5.0.58Minor · nowSep 12gald3r 5.0.58
Latest release
v5.0.58
Shipped 2026-09-12. The engine ships several releases a week; every one is listed with its date and notes.
FAQ
Questions developers ask
- What is gald3r?
- Gald3r is an AI-native task management framework that wraps around your AI coding IDE. It gives AI agents persistent memory, multi-repo orchestration, and adversarial quality gates — so they remember your architecture across sessions, coordinate changes across repositories, and cannot sign off on their own work.
- How do I install gald3r?
- Download the signed installer for Windows, macOS, or Linux (or run gald3r install update), then run gald3r setup inside your project. Gald3r is file-first — no servers or subscriptions are required for the free framework. The Quickstart walks through your first task in under 10 minutes.
- Which IDEs does gald3r support?
- 37 AI coding tools are supported. 6 have full parity — Cursor, Claude Code, Gemini CLI, OpenAI Codex, OpenCode, and GitHub Copilot — sharing identical agents, skills, commands, and task state, so you can switch between them mid-task and pick up exactly where you left off.
- Is gald3r free?
- Yes. The complete gald3r framework — all agents, skills, commands, rules, hooks, the vault, and multi-project coordination — is free under the FSL-1.1-Apache license (Functional Source License, which converts to Apache-2.0 after two years). There are no tiers. An optional self-hosted Docker backend adds MCP memory/search tools and is also free.
ᚷᚨᛚᛞᚱ
See the code. Then ship it.
Signed installers, no account required; the public templates use the FSL-1.1-Apache license posture. Install in any project in minutes and put a reviewer between your agent and your main branch.