🎉 gald3r v5.0.58 is out — gald3r 5.0.58 — choose one recommended download for your computer..See release notes →
Free · FSL-1.1-Apache · No tiers

Everything gald3r can do

One install, no feature gates. 13 agents, 141 skills, 220 commands, and a verification ladder none of them can skip.

See how it's licensed on the pricing page.

The verification ladder

Not a feature. The reason the rest of this page is trustworthy.

Every unit of work moves through the same four rungs, whether you run one task by hand or an unattended autopilot queue overnight.

  1. 01

    Implement

    An agent claims a task from the board and works it in isolation, ticking acceptance criteria with evidence — never a promise.

  2. 02

    Review, by someone else

    A different agent checks the work against the same criteria. The agent that wrote the code is structurally unable to mark it done.

  3. 03

    Verify

    Status lives in a local SQLite ledger; the markdown board regenerates from it. gald3r doctor reports honestly and never fabricates green.

  4. 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 autopilot

The suite

What's shipped since the last time this page was true

gald3r_ide — the new standalone editor

Positioned deliberately between Notepad++/UltraEdit and early VS Code — no plugin marketplace, by design.

  • 25 JSON language definitions

    Syntax, folding, and comment toggling for languages CodeMirror has no first-party pack for — PowerShell, TOON, four SQL dialects, Dockerfile, git config files, and more

  • VS Code / Notepad++ / UltraEdit keybindings

    Three presets, switchable live from the right-click menu, remembered per user

  • AI edits as diffs, per hunk

    The agent panel never writes a file directly — every proposed edit renders as a unified diff with Accept/Reject on every hunk

  • Shares Throne's editor core

    @gald3r/editor-core — one implementation, not a fork

Gald3r Throne — the desktop control plane

  • Company Admin

    Org owners/admins change member roles, remove members, invite by email, and see a billing card — plan, licenses purchased vs. assigned, usage

  • Council Room

    Convene every workspace project agent in one transcript — select/mute members, an explicit confirm, parallel fan-out, answers attributed per project

  • Workspace graph, with live activity

    Every project in the workspace manifest as a node graph — an agent-count ring per project, and a pulse traveling the edge for every new WPAC/Valkyrie message

  • Fleet

    Local model hosts (loaded models, health) beside every registry provider's token usage and quota

  • Mission Control

    Org/team command-center view of active swarms and claimed territory, with an agent-approval queue — reject requires a reason

  • IDE section

    File explorer, syntax-highlighted editor, TypeScript/JavaScript diagnostics, git-aware source control — the same per-hunk AI diff review as gald3r_ide

Longship — the agentic focus terminal

  • Full-screen TUI agent session

    Scrolling conversation with markdown/diff rendering, an approval modal for tool calls, a question modal for mid-turn agent questions, command palette

  • Free swarm dashboard

    gald3r-longship dashboard renders a live autopilot loop as a per-lane terminal view, reading a local events ledger — no world_tree login needed

  • Agent Client Protocol server

    Zed-compatible acp mode drives the same turn loop as run and ui — one implementation, three front ends

Cartograph — the code graph

  • gald3r graph explore / impact / callers

    Declarations, callers/callees, and blast radius from a pure-Go tree-sitter index — no external service, no network call

Coordination — Valkyrie vs. WPAC

  • Valkyrie (gald3r valk)

    Live HTTP connector to world_tree — polls for signed directives and answers grounded cross-project questions; requires a signed-in account

  • WPAC (gald3r workspace)

    Offline-first file transport — [CONFLICT]/[REQUEST]/[ORDER]/[INFO]/[SYNC] entries in a shared INBOX; works with no network at all

The free framework

What ships in every install, no account required

Core Framework

  • .gald3r/ project structure

    PROJECT.md, TASKS.md, BUGS.md, PLAN.md, DECISIONS.md, CONSTRAINTS.md

  • File-first (no cloud)

    Everything lives in your repo. Commits with your code.

  • FSL-1.1-Apache license posture

    Public releases are fair-source licensed; private/internal repos remain proprietary.

  • Git-native workflow

    No accounts, no databases, no telemetry

  • PROJECT.md session injection

    Agents read mission, goals, and constraints at start

  • CONSTRAINTS.md enforcement

    Hard rules agents check before marking tasks done

  • DECISIONS.md audit trail

    Append-only log of architectural choices and rationale

Task Management

  • TASKS.md master index

    Single source of truth for all work items

  • Individual task files

    tasks/ folder with spec, acceptance criteria, status history

  • Status lifecycle

    pending → in-progress → awaiting-verification → complete

  • Priority + dependency system

    Critical/High/Medium/Low; blocked-by dependencies

  • Subsystem ownership

    Tasks linked to subsystem specs for ownership tracking

  • Sprint/batch planning

    Complexity scoring and sprint load estimation

Quality Gates

  • Adversarial implement/verify split

    The agent that writes code cannot mark it done

  • @g-go-code implement command

    Structured autonomous implementation loop

  • @g-go-review verify command

    Independent verification in a fresh session

  • Acceptance criteria gate

    Agent walks every AC item before marking [🔍]

  • Bug-discovery gate

    Pre-existing bugs must be logged before task can close

  • Stub/TODO scan

    Implementation scanner blocks tasks with unresolved TODOs

Skills

  • Task management skill

    g-skl-tasks: create, update, status transitions

  • Bug tracking skill

    g-skl-bugs: BUGS.md + bugs/ lifecycle

  • Feature management skill

    g-skl-features: staging → specced → committed → shipped

  • Code review skill

    g-skl-code-review: security, quality, performance review

  • Git commit skill

    g-skl-git-commit: conventional commits with task refs

  • Project health skill

    g-skl-medic: tiered L1-L4 .gald3r/ health diagnostics and heals

  • Status skill

    g-skl-status: session context, active tasks, phase

  • Recon & research suite

    g-skl-recon-repo/url/docs/yt/file + res-review → res-deep → res-apply

  • Knowledge vault + ingestion

    Obsidian-compatible vault, platform-doc crawl, URL and YouTube ingestion

Multi-Project Coordination (WPAC)

  • Project topology linking

    Parent/child/sibling project registration

  • Cross-project task ordering

    @g-wpac-order: push tasks from parent to child

  • Child-to-parent requests

    @g-wpac-ask: blocked tasks request parent action

  • FYI notifications

    @g-wpac-notify: lightweight cross-project messaging

  • Sibling contract sync

    @g-wpac-sync: advisory peer-to-peer coordination

  • Spawn child project

    @g-wpac-spawn: create new gald3r project linked to parent

  • File/asset transfer

    @g-wpac-move / @g-wpac-send-to: cross-repo asset migration

Hook System

  • Event hooks

    Trigger scripts on agent events: session start, task done, commit

  • Pre-built hooks

    g-hk-setup-user, g-hk-validate-shell, and the shared Python hook core

Optional Self-Hosted MCP Backend

Opt-in and free — run it yourself with Docker for MCP-backed memory. The framework is fully functional without it.

  • 16 MCP tools

    Unified server exposed to all AI agents via Docker

  • Agent memory — semantic search

    memory_search: vector search across all sessions

  • Agent memory — session capture

    memory_capture_session: AI self-reports summaries

  • MediaWiki integration

    mediawiki_page, mediawiki_search: full wiki R/W

  • Oracle DB tools

    oracle_query, oracle_execute: source + target DB access

  • Platform docs search

    platform_docs_search: crawled Cursor/Claude/Gemini docs

  • Project health scoring

    gald3r_health_report: 0–100 score with per-subsystem breakdown

  • Docker Compose runtime

    Runs locally, no cloud required

IDE Support

  • 37 AI coding tools

    One install activates gald3r across every supported tool — Cursor, Claude Code, Gemini, Codex, OpenCode, Copilot, Zed, Kimi Code, and 29 more

  • 6 at full parity

    Cursor, Claude Code, Gemini, Codex, OpenCode, and GitHub Copilot share identical agents, skills, commands, and task state

  • Same skills/agents/commands across IDEs

    Install gald3r once, use everywhere — .cursor/, .claude/, .codex/, .gemini/, .opencode/ parity maintained

  • Cross-IDE session handoff

    Start in Cursor, continue in Claude Code — state lives in .gald3r/, not the IDE

Every binary is free. Clone the repo and put a reviewer between your agent and your main branch.