Concepts

What is Galdr?

Galdr is the infrastructure layer between you and your AI coding tools. It gives agents durable memory, cross-repo coordination, and adversarial quality enforcement — without requiring you to change how you code.

The Three Problems

01 The Memory Problem

Session 1: You explain your architecture — PostgreSQL not SQLite, REST not GraphQL, no ORM.
Session 12:The agent suggests Prisma with SQLite. It doesn't remember session 1.

Every AI conversation starts from zero. Your constraints, conventions, and architectural decisions are forgotten the moment you close the chat.

Galdr's answer: CONSTRAINTS.md + learned-facts.md

Every decision gets written to files that agents load at session start. Constraints are non-negotiable — any violation is flagged and blocks task completion. Learned facts capture conventions, preferences, and context that accumulate over time.

02 The Multi-Repo Blindness Problem

You're running 4 repos: api, web, mobile, shared-lib.

A breaking change lands in shared-lib. The agent in the web repo has no idea. It keeps building against the old API. By the time you notice, the divergence is deep.

Galdr's answer: PCAC (Parent-Child-Agent-Coordination)

Declare a parent/child/sibling topology between your repos. When a breaking change happens in shared-lib, broadcast a task to all downstream repos with @g-pcac-order. Each child receives it in their INBOX.md and can act independently — or block and ask the parent for help.

03 The Self-Review Bias Problem

You ask the agent that wrote the code to verify it. It passes everything — of course it does. The bugs it introduced are invisible to it. The edge cases it missed are the same ones it would miss when verifying.

Galdr's answer: Adversarial Two-Phase Review

The agent that implements code can only mark it [🔍] (awaiting verification). It is structurally incapable of marking it [✅]. A separate agent session runs @g-go-review and checks the work independently. Verification failure sends the task back to pending with a FAIL note.

How Galdr Works

Galdr is entirely file-based. It installs as a set of directories (.galdr/, .cursor/, .claude/, etc.) at your project root. Agents load these files at session start and write to them as they work.

.galdr/PROJECT.md

Mission, goals, and project identity. Loaded at every session start.

.galdr/TASKS.md

Master task index with status. Individual task files in .galdr/tasks/.

.galdr/CONSTRAINTS.md

Non-negotiable architectural rules. Checked before any task completes.

.galdr/BUGS.md

Bug index with severity. Critical and high bugs run before new features.

.galdr/DECISIONS.md

Append-only log of architectural choices and rationale.

.galdr/PLAN.md

Milestones and delivery strategy.

Before and After Galdr

Scenario	Without Galdr	With Galdr
New session starts	Agent asks: 'what are we working on?'	Agent loads context: goals, active tasks, constraints
Task verification	Same agent marks own work done	Separate session independently checks criteria
Breaking API change	Downstream repos break silently	PCAC order broadcast to all children; each gets INBOX entry
Switching IDEs	Lose all context, start over	Same .galdr/ files; identical state in any IDE
Tech stack drift	Agent suggests random libraries	CONSTRAINTS.md enforced; any violation blocked
Multiple repos	No coordination between AI sessions	Parent/child topology; task cascade; INBOX routing

Ready to try it?

Get Started → 5 minutes Quickstart guide