factory
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@factoryrequest a child task in billing to fix invoice rounding"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
factory
Kanıt-tabanlı gate + domain denetleyicisi — software factory kontrol katmanı.
Multica (veya herhangi bir runner) üstünde, "task bitti" iddiasını kanıta
bağlayan ince katman: analiz → kod → bağımsız review → gerçek test → context
→ contract sync. Hepsi fail-closed; kanıt yoksa done yok.
Mimari ve kararlar: docs/ARCHITECTURE.md, docs/adr/.
Kurulum
uv sync # dev: uv run pytest -q
uv pip install . # ya da: pip install .Related MCP server: permitd MCP Server
Hızlı başlangıç
WS=./factory-workspace
factory $WS domain add --name auth --profiles coder-auth
factory $WS domain add --name billing --profiles coder-bill --depends-on auth --critical
factory $WS task add --domain auth --title "JWT rotasyonu"
factory $WS claim --worker w1 --domain auth
# teslim zinciri (sırasıyla; hepsi kanıt bırakır):
factory $WS check 1 # gate raporu; geçmeyense exit 1
factory $WS test-run 1 -- pytest -q # testi GERÇEKTEN koşar, raporlar
factory $WS review 1 --reviewer r-x --coder coder-auth --approve \
--artifact $WS/03-review.md
factory $WS context 1 --path domains/auth/CONTEXT.md
factory $WS finalize 1 # gate'ler geçmeden done OLAMAZGate'ler (fail-closed)
Gate | Kanıt |
analysis_done | 01-analysis.md var ve boş değil |
reviewer_approved | farklı profilden APPROVE; artefakt yerinde; kritik domainde 2 bağımsız onay; kod digest'i değişince bayatlar |
tests_passed | gerçek koşu + exit 0 + rapor dosyası; rapor hash'i kayıtla birebir |
context_updated | CONTEXT.md güncellendi ya da reviewer onaylı muafiyet |
contract_sync | contract'a dokunulduysa tüm bağımlı domain inceleme task'ları done |
Deploy
Docker:
docker build -f deploy/docker/Dockerfile -t factory .
docker compose -f deploy/docker/docker-compose.yml upKubernetes (tek-yazar SQLite → gatekeeper replicas=1; worker'lar Job olarak
factory /data check çalıştırır):
kubectl apply -f deploy/k8s/factory.yamlMultica entegrasyonu (plan)
Multica issue review state'ine geçtiğinde webhook/loop bu repo'nun
check+finalize çağrısını çalıştırır; exit 0 olmadan issue done olmaz.
Worker'lar Claude Code/Codex CLI; kanıt dosyaları task workspace'inde durur.
Ajan araçları (birbirini çağırma)
Ajanlar serbest chat yapmaz; registry-denetimli araçları çağırır (ADR-006):
request_task (başka domain'e çocuk iş), report_result, ask_domain
(hedef domain'in CONTEXT.md+contract'ları), request_contract_change
(sahibi+aktif worker), submit_plan, approve_plan. Kimlik = aktif claim.
Claude Code'a MCP olarak bağlanır:
claude mcp add factory -- python -m factory.mcp --workspace $WSCLI'dan da aynı dispatch: factory $WS call request_task '{"caller":...}'
Araç listesi: factory $WS tools.
Lead beyni (plan→onay→dispatch→gate sweep):
factory-lead $WS --runner 'claude -p' --once # veya --interval ile döngüMultica köprüsü (UI = Multica, kanıt = Factory)
Factory task'larını Multica issue'larına bağla, sonra döngüyü sürdür:
link_issue(store, task_id, "<multica-issue-uuid>") # veya meta'ya elle
MULTICA_TOKEN=*** MULTICA_WORKSPACE_ID=... \
python -m factory.multica $WS --interval 30Gate'ler yeşilse issue in_review olur; kanıt kaybolursa geri çekilir;
sen UI'da done dersen Factory task'ı kapanır (kanıtsızsa gate_bypass
denet olayıyla). Multica'yı upstream'e bağlı tutmak için:
scripts/multica-follow.sh (ADR-007: fork yok, API köprüsü).
Test
uv run pytest -q # 38 test: gate semantiği, claim/lease, domain, DAGThis server cannot be deployed
Maintenance
Related MCP Connectors
Runtime permission, approval, and audit layer for AI agent tool execution.
Context and control for any AI. Agent identity, scoped access, and proof across tools.
Agent routing, safety preflight, execution receipts, best execution and verified outcomes.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to execute tools securely with Docker isolation, tier-based permissions, and feature flags, providing a centralized tool registry and MCP server.-
- AlicenseNot gradedqualityBmaintenanceGates agent tool execution with human approval, audit trails, and replay-resistant permits, enabling safe use of tools in agent loops.MIT
- FlicenseNot gradedqualityCmaintenanceEnables controlled AI-agent access to enterprise-shaped tools with a deny-by-default gated write path, human approval, dry-run execution, and append-only audit logging.1-
- FlicenseNot gradedqualityCmaintenanceEnables agents to securely discover and invoke a centrally governed catalog of tools from distributed internal and external providers, with policy enforcement, quotas, inspection, and audit controls.-