Node palette (left)
The capabilities you can add — Observe, Source, Transform, Validate, Decide, Approval, Publish and domain nodes — dynamically reflecting what the current runtime supports.
Forge Studio
Compose agents
visually.
Forge Studio is the visual composer where production experts design, inspect, test and publish agents in domain language — no hand-editing JSON. Studio is not a runtime: it converts agent JSON to a graph and back losslessly, and preserves fields it cannot draw.
1
Layout
The Studio interface
Canvas & pipeline spine
The agent as a graph: drag to add, connect stages, insert or delete nodes, and repair a broken flow. The spine shows execution order.
Inspector (right)
Per-node configuration: label, type, execution mode, inputs/outputs, policies, validators and evidence — tabbed and contract-aware.
2
Modes
Manual, AI & automatic execution
Manual
A human performs the step; Forge records it and hands off honestly. Used where no safe automated executor exists yet.
AI-assisted
Fabric performs the step through a real capability (vision QA, grounded drafting, reasoning), always grounded and confidence-scored.
Automatic
A deterministic executor runs the step (Photoshop JSX, Unity bridge, Perforce sync, data write) — only if the executor registry supports its mode and runtime.
“AI everywhere except sign-offs.”
AI is the default execution mode wherever a genuine capability exists — but it never fabricates success and never replaces the human approval gate. A step with no safe executor is an honest manual hand-off, not a faked “done”.
3
Workflow
Authoring an agent
Start from blank, a clone, or Agent Fabricator
Agent Fabricator turns a ticket or plain-language goal into a grounded candidate graph with typed contracts, capability bindings and review evidence. You review the proposal; Fabric never auto-publishes it.
Configure nodes & connect stages
Set each node’s type, execution mode and config; wire the spine. A flow-break check catches disconnected or invalid graphs.
Lint & simulate
Structural lint plus a Harness simulation produces an explainable readiness result before anything runs for real.
Save as Draft, then Publish
Publishing makes the definition live; it does not execute it. Versions provide history, visual diff and rollback.
4
Trust
Preview, validation & certification
Preview verifies capability-to-executor coverage and shows proposed reads and writes. Simulation runs the Harness checks and produces a readiness certificate with a confidence meter. A high score never makes an unsupported or outward action safe — executor, policy, validation and approval gates remain binding, and the Harness is advisory to the approval gate, never a hidden blocker. Golden baselines let you catch regressions (▲/▼) against a known-good certification.
The certification stack
Beyond the readiness certificate: golden baselines + regression, security checks (secrets, shell-safety), replay (predicted vs actual), contract coverage + the AUTO-executor gate, and the Fleet Quality Gate — a studio-wide, weakest-first roll-up that certifies the whole roster before any run is trusted.
Intelligence Profiles & sealed skills
Intelligence Profiles v1.2.0 prove an agent's readiness across workspaces — resolved capabilities, installed/enabled knowledge, sealed runtime skill manifests, and required-tool readiness. Skills (48 from .claude/skills) follow a knowledge-vs-executable contract: RAG-ready guidance immediately, execution bound and sealed by version + digest at run start — no silent drift.
5
Reuse & recovery
Templates, components, versioning & troubleshooting
Templates & components
Start new agents from templates; reuse proven node groups as components. Treat a component like shared production code — narrow contracts, versioned changes, retest every consumer.
Versioning & diff
Draft → Published with a visual diff and rollback, so a change is reviewable and reversible.
Common issues
Flow breaks (use Repair flow), unsupported AUTO stages (pick a supported mode/executor), low source coverage (add grounding sources).
When Agent Fabricator cannot compile a goal
Inspect its stated missing requirement or capability gap. Fabric retries malformed model output through governed planning and deterministic compilation, but it will not invent a tool or claim an unsupported node is executable. Check model health, provide authoritative inputs, or bind a qualified capability.
Upgrade Agent
Modernizes a legacy agent as a preview: preserves authored behavior, proposes typed capabilities and executor bindings, and exposes every unresolved physical stage. Apply to canvas remains undoable and does not save, publish, commission or run the agent.
Retire legacy behavior only after proof
Upgrade, Check, Test and Certify each agent, then complete a supervised production pilot with retained destination and outcome evidence. Keep the legacy runtime available until every dependent agent has passed that sequence; migration is per agent, never a bulk trust assumption.
6
Federation
Federated reuse & coordinated migration
Publishing a callable agent lets other workspaces import a governed, credential-safe copy that stays pinned to the exact source version it was taken from. When the source owner changes a shared dependency, Studio does not silently mutate those copies — it helps the owner drive every affected copy through one sealed migration campaign, so a fleet-wide upgrade stays reviewable, evidence-bound and reversible.
Sealed migration campaigns
The copies impacted by one dependency change are grouped into a single durable campaign bound to that exact impact set. Each draft moves forward through prepared → reviewing → certified → published (failed is an off-path terminal), every transition appends immutable evidence, and a completion roll-up reports the campaign as in-progress, needs-attention or complete.
Governed contract remapping
If the upgrade changes the dependency's input/output contract, the upgrade is blocked until you map the boundary. Remap-preview auto-maps same-name, same-type fields and lists exactly which inputs and outputs still need an explicit mapping or default; the reviewed plan is digest-bound and travels with the draft.
Harness-bound certification
A draft only reaches certified when the real Harness certifies that agent — Studio never trusts a caller-supplied state for that step. Manual advances are limited to reviewing or failed; certified and published are written by governed gates and the publish path.
Durable notifications & reconciliation
Consumers of a changed dependency receive retryable cross-workspace notices tracked in a durable delivery ledger: pending → delivered, or failed with bounded retry and dead-lettering after five attempts. Reconciliation re-attempts every outstanding notice, so a transient outage never drops a consumer.
Agent Store & portable packages
The Agent Store is a searchable catalog of commissioned, available and decommissioned agents — showing origin, execution model, capabilities, contracts, success rate and certification. Export a checksum-sealed forge.agent-package.v1 bundle (portable graph, profile metadata, permitted local skills; credential-shaped fields stripped). Preview is a bounded read; installation is a separate admin-gated action that never grants credentials, tool access or commissioning authority.
FABRIC-origin & guided release
Agents created or materially upgraded by Fabric carry the FABRIC origin treatment, with Harness-bound certification shown separately. The guided lifecycle — Build → Check → Test → Resolve → Certify → Save → Publish → Commission — keeps uncommissioned pilots preview-only, with no real production writes until commissioned.