Multi-Agent LLM Coordination Framework

February 27, 2026 · View on GitHub

A practical methodology for coordinating multiple AI agents on complex software projects.

Built and proven on an 8,521+ test, 338-formula deterministic game engine — by a non-technical operator (former chef, English educator) coordinating Claude Opus, Claude Sonnet, and other LLM agents with zero shared memory across 100+ agent sessions over 4+ months.


The Problem

When you use multiple LLM agents to build software, you hit walls that don't exist in solo development:

  • Agents forget everything between sessions. Every new context window starts from zero.
  • Parallel agents conflict. Two sessions editing the same file produce merge disasters.
  • Nobody holds the full picture. No single agent (or human) can keep the entire project in context.
  • Documents drift from reality. Agent-written docs become stale, contradictory, and self-reinforcing.
  • The human coordinator becomes the bottleneck. You can't read everything, and you can't tell which agent is right when they disagree.

These aren't theoretical problems. They're what happens on Day 3 of any serious multi-agent project. This framework documents the solutions we discovered by breaking things and codifying the fixes.


What Makes This Different

Most AI coordination guides tell you what to do. This one shows you what went wrong.

From the proving ground (D&D 3.5e combat engine):

  • 8,521+ automated tests across 9 verification domains
  • 338 formulas verified against source rules
  • 30 bugs found and categorized into 12 coordination error patterns
  • 100+ work orders dispatched across 100+ agent sessions over 4+ months
  • 3 of 7 parallel agents silently failed to commit in one dispatch — the fix became a governance pattern

Every pattern in this framework exists because something specific broke. The failure catalog has the receipts.


Who This Is For

  • Non-technical builders using AI agents to construct software ("vibe coders")
  • Developers scaling from one AI assistant to a fleet of specialized agents
  • Researchers studying multi-agent LLM coordination
  • Anyone who has had an AI agent confidently break something another agent just fixed

Core Principles

1. Artifact Primacy

If it's not in a file, it doesn't exist.

Agents have perfect recall within a context window and zero recall across context boundaries. Conversational knowledge vanishes at session end. Every fact, decision, and status that must survive a context rotation must be pinned to a file.

Implication: Your project's files aren't just code and docs — they're the only shared memory your agent fleet has.

2. Staged Context Loading

Reading order matters more than reading volume.

Agents waste context window space when they read files in the wrong order or read everything at once. A defined reading sequence — orientation first, then state, then rules — gets agents operational in minimal context.

Pattern: Compass (what is this?) → State (what's done?) → Rules (how do we work?) → Task (what do I do?)

3. Dispatch Self-Containment

Every work order must be executable by an agent with zero prior context.

The work order plus the files it references must contain everything an agent needs. No reliance on "the previous agent will have explained this." Test: could a brand-new agent execute this dispatch using only the dispatch file and linked references?

4. Machine Truth Over Prose Truth

When a script and a document disagree, the script is right.

Agent-written prose drifts from reality within 3-4 handoffs. Machine-generated state (test counts, git status, automated snapshots) doesn't. Build scripts that produce canonical facts, and treat everything else as commentary.

5. Protocol Over Memory

Solve coordination with protocols, not shared state.

You can't give agents shared memory. You can give them protocols — handoff checklists, consistency gates, session scope declarations, structured memo formats. The protocols are the coordination mechanism.


Framework Components

Patterns

Reusable solutions to specific coordination problems. Each pattern documents the problem, the solution, when to use it, and a real example from the proving ground.

PatternProblem It SolvesSolution
Enforcement HierarchyCorrections don't stick across sessions3-tier model: test-enforced > process-enforced > prose-enforced
Staged Context LoadingAgents waste context reading files in wrong orderDefined reading sequence that orients agents in minimal context
Dispatch Self-ContainmentWork orders fail because agents lack contextSelf-contained work orders that amnesiac agents can execute
Artifact PrimacyKnowledge lost between sessionsPin every decision to a file — conversation doesn't survive
Cross-File Consistency GatePartial updates create contradictionsAll-or-nothing updates across every file referencing a fact
Session BootstrapAgents start with stale assumptionsMachine-verified truth (tests, git status) before prose documents
Concurrent Session ProtocolParallel sessions conflict on shared filesExplicit file ownership with no overlaps between agents
PM Context CompressionHuman coordinator can't read everythingStructured information flow to a bandwidth-limited human
Coordination Failure TaxonomySame mistakes repeat across projectsCategorized catalog of what goes wrong and why
Role SeparationAgents assume they can do everythingFive-role model with explicit authorities and boundaries
Research-to-Build PipelineRaw insights produce scope bleed when dispatched directlyStaged conversion: Burst → Research → Brick → Builder WO
Plain English PassNon-technical operators can't parse technical debriefs3-question translation layer before the technical dump
Debrief Integrity BoundaryAgent self-reports are trusted without verificationNamed trust boundary with verification spectrum and mitigations
Integration CanaryUnit tests pass but product doesn't work end-to-endOne script that exercises the full product path before new WOs
Parallel Implementation ParitySame logic implemented in multiple paths drifts silentlyEnumerate all parallel paths on every WO; verify parity before debrief
Sweep Audit ProtocolSequential WOs can't see system-wide coherence driftPeriodic read-only audits that cross-check subsystems for consistency
Authority TaggingCommunity convention ships instead of specificationEvery domain-logic WO declares SPEC or POLICY authority — no third type
Post-Debrief RetrospectiveBuilder peripheral observations die in context windowsMandatory post-debrief question surfaces what the builder noticed
Hidden Assumption SweepSmall questions that reframe architecture go unclassified10-question triage protocol converts grenades into named artifacts
Worktree Isolation ProtocolParallel agents collide at filesystem level despite non-overlapping file scopeOne git worktree per active builder — isolation at the OS level, not just the work order level
Proactive Assumption Sweep CadenceReactive sweeps miss confidently-wrong assumptionsScheduled 10-question sweeps on a fixed cadence — pair with every audit dispatch

Templates

Ready-to-use file templates for implementing the patterns in your own project.

TemplatePurposeFile
Onboarding ChecklistReading order + verification steps for new agentsImplements Staged Context Loading
Work Order DispatchSelf-contained task assignmentImplements Dispatch Self-Containment
Audit Work OrderRead-only audit task assignment (never writes code)Implements Sweep Audit Protocol
Handoff DocumentEnd-of-session knowledge transferImplements Artifact Primacy
Session MemoStructured report to human coordinatorImplements PM Context Compression
Sources of Truth IndexWhich file is authoritative for which conceptImplements Machine Truth Over Prose

Case Study

How these patterns were discovered and proven on a real project.

DocumentContent
Project OverviewWhat was built, by whom, scale and complexity
Failure CatalogReal coordination failures, categorized with root causes
MetricsQuantitative results — formulas verified, bugs found, reclassification rates
Lessons LearnedWhat worked, what didn't, what we'd do differently

Quick Start

If you're non-technical and want to learn this methodology from scratch, start with the Methodology Guide — a step-by-step walkthrough written for someone who has a problem, has access to AI agents, and has no idea how to start.

If you're already running agents and want to add specific coordination patterns, start here:

  1. Create an onboarding checklist (template) — this is the single highest-impact intervention. Define what agents read, in what order.

  2. Create a sources of truth index (template) — before your second session, decide which file is authoritative for each concept.

  3. Write self-contained work orders (template) — the moment you start dispatching tasks to agents, the dispatch must stand alone.

  4. Add a session bootstrap script (pattern) — even a simple script that prints git status + test count grounds the agent in reality instead of potentially-stale docs.

  5. Use handoff documents (template) — when a session ends, write what the next agent needs to know. This is not optional.


Origin

This framework wasn't designed top-down. It was discovered bottom-up by a non-technical operator coordinating multiple AI agents (Claude Opus, Claude Sonnet, GPT-4) to build a D&D 3.5e deterministic referee engine. Every pattern exists because something broke and the fix got codified into a protocol.

Project stats (updated 2026-02-27):

  • 8,521+ automated tests (zero regressions across all accepted batches)
  • 338 formulas verified against source rules across 9 domains
  • 30 bugs found and categorized into 12 error patterns
  • 100+ work orders dispatched (fix, feature, research, governance, audit, framework)
  • 50+ builder debriefs archived
  • 30 research documents produced
  • 100+ agent sessions coordinated through file-based protocols
  • 25+ delivery batches, each with gate tests as the acceptance arbiter
  • Every governance document born from a specific, documented failure

The operator's background: Former chef, current English educator, zero programming experience. The methodology emerged because protocols were the only tool available — and it turns out protocols are the right tool for coordinating agents that can't remember yesterday.


Contributing

This is a living framework. If you're using these patterns on your own multi-agent project, your failure cases and solutions make the framework better. Open an issue with:

  • What pattern you tried
  • What worked or didn't
  • What coordination failure you hit that isn't in the taxonomy

License

MIT License — use freely, adapt to your projects, share what you learn.


"Creative in voice, strict in truth, accountable in every outcome."