convez.ai
AgenteDesenvolvimento WebAvançado
DammianMiller

Universal Agent Protocol (UAP)

DammianMiller/universal-agent-protocol

Memória, coordenação e gates executáveis para agentes IA disciplinados

Ver no GitHub
InicialPluginScore · de 0 a 100

Framework de produção que adiciona memória persistente, julgamento e disciplina a agentes IA de codificação. Implementa um protocolo de convergência (`uap deliver`) que itera modelos contra gates reais de compilação, testes e lint até entregar trabalho verificado, reduzindo tokens e erros em até 98%.

Instalação

$npx convez-ai DammianMiller/universal-agent-protocol

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 1 dia
Versão
v2.52.2
Issues abertas
1
Contribuidores
3

Casos de uso

  • Manter contexto e aprendizado entre sessões de codificação com agentes IA
  • Garantir qualidade de código com gates automáticos (build, testes, lint)
  • Coordenar múltiplos agentes em repositórios compartilhados sem conflitos
  • Reduzir consumo de tokens com compressão inteligente de outputs

Veredito

UAP é uma camada de produção robusta e bem-testada para engenheiros que precisam de agentes confiáveis em pipelines reais. A abordagem de gates executáveis e memória semântica resolve problemas genuínos de AI coding, mas requer TypeScript/Node e integração profunda no projeto.

README

A tradução para o português deste README ainda está na fila.

Universal Agent Protocol (UAP)

Give your AI coding agents memory, judgment, and the discipline to finish the job.

npm Node Tests License

v1.40.0 · 168 modules · 117 test suites · 9 agent harnesses

Quickstart · Why UAP? · uap deliver · Architecture · Benchmarks · Docs


Why UAP?

AI coding agents are capable but undisciplined. They forget everything between sessions, burn tokens echoing huge tool outputs, repeat the same mistakes, declare victory on work that doesn't compile, and trip over each other in shared repos. UAP is a production-tested layer that sits underneath your agent harness (Claude Code, Factory, Cursor, OpenCode, and more) and fixes these problems at the protocol level — no model change required.

The problemWhat UAP doesMeasured impact
Agents forget past sessions4-tier memory with semantic recall + write-gates49.7% fewer tokens
Tool output floods the contextMCP Router — tool-hiding + FTS5 output compressionup to ~98% on large tool calls
Agents declare done on broken workuap deliver — convergence loop against real gates+33pp task success (25% → 58%)
Repetitive mistakes23 Terminal-Bench patterns + learning loop68% fewer errors
Wrong model for the jobMulti-model router, 7 profilesoptimal cost/perf per task
Agents step on each otherWorktree isolation + coordination serviceconflict-free parallel work
"Guidelines" get ignoredPolicy gates as executable hooks, not proseviolations are blocked, not suggested

Benchmarks below are from Terminal-Bench 2.0 (12 representative tasks). See docs/benchmarks/ for the full methodology and raw data.


Quickstart

# Install globally
npm install -g @miller-tech/uap

# One-command setup in your project (memory, patterns, hooks, policies)
cd your-project
uap setup

That's it. Your agent now has persistent memory, battle-tested patterns, policy gates, and multi-agent coordination wired into every session.

uap memory query "how did we handle auth last time?"   # semantic recall
uap deliver "add rate limiting to the API"             # drive a model to verified completion
uap dashboard overview                                  # live task / agent / memory state

The deliver harness

uap deliver is the headline of the v1.27–v1.40 line: a convergence loop that iterates a model against your project's real completion gates until the work is actually delivered — build passes, tests pass, lint is clean — not until the model thinks it's done.

uap deliver "implement the password reset flow"

What happens under the hood:

  1. Explore → plan → apply — the model proposes changes; the applier writes them safely (pre-existing tests and gate configs are protected from being overwritten).
  2. Verify against real gates — a verifier ladder runs your build, tests, and lint. Nothing is "done" until they're green.
  3. Critique & iterate — failures feed back as structured guidance; the loop continues, persisting until delivered (extends past --max-turns to a ceiling, stopping on genuine stagnation).
  4. Auto-optimization — every task is classified by complexity and the matching aids (HALO trace analysis, divergent ideation, coordination, deploy batching) activate automatically.
  5. Autonomy with a guidance channel — runs the full mission without stopping to ask, while still accepting operator guidance mid-flight.

It works with frontier models and local models (llama.cpp / Qwen) served over the Anthropic Messages API. See docs/guides/DELIVER.md.


Features

  • 🧠 4-tier memory — daily log → working cache → semantic (Qdrant) → long-term archive, with write-gates that block low-quality/duplicate memories and corrections that cascade across tiers.
  • 🗜️ MCP Router — a token-optimizing tool proxy; large outputs are compressed via FTS5 intent search instead of dumped into context.
  • 🎯 uap deliver — the convergence/delivery harness (above).
  • 🌳 Worktree workflow — isolated branch-per-feature, auto-PR, safe cleanup; enforced so agents never edit the project root.
  • 🛡️ Policy gates — 20 executable enforcers (worktree, test, schema-diff, expert-review, memory-before-plan, delivery-enforcement…) that block non-compliant tool calls.
  • 🤖 Expert droids & skills — 38 specialized droids and 32 skills, with an expert-router that recommends a droid chain per task.
  • 🧭 Multi-model routing — 7 profiles (Claude Opus/Sonnet/Haiku, GPT, Qwen, generic); the router picks by complexity, cost, and performance.
  • 🚦 Deploy batching & coordination — batched git/deploy actions and overlap detection keep multi-agent work conflict-free.
  • 📊 Dashboard — rich TUI/web views of tasks, agents, memory, benchmarks, and policy status.
  • 🔌 9 harnesses — Claude Code, Factory, Cursor, VSCode, OpenCode, Codex, ForgeCode, Oh-My-Pi, Hermes.

Full list with code-level detail: docs/reference/FEATURES.md.


Architecture

UAP installs hooks into your agent harness, then mediates every tool call through memory, policy, and token-optimization layers.

┌─────────────────────────────────────────────────────────────┐
│  Agent harnesses                                            │
│  Claude Code · Factory · Cursor · VSCode · OpenCode · …     │
└───────────────────────────┬─────────────────────────────────┘
                            │ hooks (PreToolUse / tool.execute.before)
                            ▼
┌─────────────────────────────────────────────────────────────┐
│                       UAP CLI (uap)                         │
│  setup · memory · deliver · worktree · policy · deploy      │
│  task · droids · model · mcp-router · harness · ideate …    │
└──┬─────────┬──────────┬──────────┬──────────┬───────────────┘
   ▼         ▼          ▼          ▼          ▼
 Memory   Policy    MCP Router   Delivery   Coordination
 4 tiers  20 gates  FTS5 compr.  harness    + deploy batch

Benchmarks

Terminal-Bench 2.0, 12 representative tasks, UAP-on vs. baseline:

MetricBaselineWith UAPΔ
Tokens consumed558,000280,438−49.7%
Task success rate25%58%+33pp
Errors per task1.170.42−68%
Wall-clock (total)618s266s−57%

Methodology, raw runs, and cost analysis: docs/benchmarks/.


Supported harnesses

HarnessHooksMCP RouterPolicy gates
Claude Code✅✅✅
Factory✅✅✅
Cursor✅✅✅
VSCode✅✅✅
OpenCode✅✅✅
Codex✅✅✅
ForgeCode✅✅✅
Oh-My-Pi✅✅✅
Hermes (global)✅✅✅

Install into all detected harnesses with uap hooks install; audit coverage with uap hooks doctor. Matrix: docs/reference/PLATFORMS.md.


Documentation

Getting StartedInstallation, quickstart, configuration
Guidesdeliver, memory, MCP router, worktrees, policies, multi-model, local models
ArchitectureSystem overview + the UAP protocol
ReferenceCLI, API, patterns, database schema, platforms
BenchmarksMethodology and results
ContributingDev setup, gates, conventions

Start at the documentation index.


Testing

npm install
npm run build      # TypeScript compile
npm test           # vitest — 117 suites
npm run bench      # benchmark suite

License

MIT © Miller Tech. See LICENSE.

Ver o original no GitHub

Bom para

Relacionados