Team Workflows
How PMs and developers collaborate in Forge — and the different ways to configure it for your team.
Forge is built for teamwork. The PM brings the vision. The developer brings the technical reality. Forge guides the conversation between them and produces a contract both sides trust.
This page explains the collaboration flow, the roles involved, and the different configurations you can choose depending on how your team works.
The PM–Developer Flow
Here's what a typical ticket looks like when both PM and developer are involved:
PM creates ticket
"Users should be able to filter by tags"
│
▼
Forge asks clarification questions
PM answers in plain language
│
▼
Spec is generated (Draft)
Acceptance criteria, file changes, wireframes
│
▼
PM assigns a developer
│
▼
Developer reviews via CLI
Adds technical context:
"Use the existing Zustand store pattern"
"Filter server-side for performance"
"Reuse the TagBadge component"
│
▼
PM reads developer feedback
Optionally re-bakes the spec
│
▼
PM approves → AEC is forged
Both sides agreed. Contract is locked.
│
▼
Developer (or AI agent) executes
Full context. No ambiguity.
This is the full lifecycle — and it produces the best results. But Forge is flexible. Not every team works this way, and not every ticket needs the full cycle.
Configuration Options
Forge adapts to your team's constraints. Here are the main ways to use it:
At a Glance
| Configuration | Codebase connected | Developer review | AI agent execution | Best for |
|---|---|---|---|---|
| Full lifecycle | Yes | Yes | Yes | Teams that want maximum spec quality and AI-assisted implementation |
| Full lifecycle, no agent | Yes | Yes | No | Teams where developers implement manually but want verified specs |
| PM + Developer, no codebase | No | Yes | No | Companies that can't connect GitHub (security policies, air-gapped repos) |
| PM only | Optional | No | No | Solo PMs, early-stage teams, or quick drafts that don't need developer review |
Full Lifecycle (Recommended)
The complete experience. Connect a GitHub repo, generate code-aware specs, have the developer review, and execute with AI assistance.
Who does what:
| Step | Who | What happens |
|---|---|---|
| Create ticket | PM | Describes the feature, connects a repo, answers clarification questions |
| Generate spec | Forge | Produces acceptance criteria, file changes, API contracts, wireframes — all referencing real code |
| Review | Developer | Reviews the spec via CLI, adds technical context and codebase knowledge |
| Re-bake | Forge | Updates the spec with the developer's input |
| Approve | PM | Locks the contract — both sides agree on what "done" means |
| Execute | Developer + AI | Developer runs forge develop, AI agent implements with full context |
Result: The highest quality specs. The developer's real-world knowledge is baked into the contract. The AI agent builds the right thing first time.
Full Lifecycle, No Agent
Same as above, but the developer implements manually instead of using an AI agent. The spec is still code-aware and developer-reviewed — the developer just writes the code themselves.
When to use: Teams where AI coding agents aren't adopted yet, or for complex work that needs a human touch.
PM + Developer, No Codebase
Some companies can't connect their GitHub repositories — security policies, compliance requirements, private infrastructure, or non-GitHub hosting (GitLab, Bitbucket, self-hosted).
Forge still works. The PM creates the ticket without connecting a repo. The spec is generated from the description and clarification answers alone — it won't reference specific files, but it still produces structured acceptance criteria, scope boundaries, and a clear contract.
The developer makes it real: When the developer reviews via the CLI, they add the codebase context manually — which files to touch, which patterns to follow, which code to reuse. The re-baked spec then incorporates this knowledge.
Who does what:
| Step | Who | What happens |
|---|---|---|
| Create ticket | PM | Describes the feature, answers clarification questions (no repo connected) |
| Generate spec | Forge | Produces acceptance criteria, scope, API contracts — without file-level references |
| Review | Developer | Adds the missing codebase context: file paths, patterns, edge cases |
| Re-bake | Forge | Updates the spec with the developer's codebase knowledge |
| Approve | PM | Locks the contract |
| Execute | Developer | Implements using the verified spec as their guide |
Result: You still get the structured protocol, the mutual agreement, and the scope boundaries. The spec is less precise on file changes, but the developer review fills that gap.
PM Only
For quick drafts, solo PMs, or early-stage teams without a dedicated developer review process. The PM creates and enriches the ticket entirely from the web app.
When to use:
- Early exploration — drafting specs before a developer is assigned
- Small teams where the PM has technical context
- Quick tickets that don't justify the full review cycle
The developer can always be brought in later. A PM-only ticket can transition to the full lifecycle at any point.
The Developer's Perspective
For developers, Forge changes the experience from "figure out what the PM meant" to "review a clear proposal and add your expertise."
What developers do in Forge
-
Review the spec — Read the generated acceptance criteria, file changes, and scope. Is anything missing? Unrealistic? Overcomplicated?
-
Add technical context — Answer questions from the AI:
- "Which existing components should be reused?"
- "What's the best approach for state management here?"
- "Are there edge cases the PM might not know about?"
-
Sign off — The developer's review is visible to the PM. Once the PM re-bakes and approves, the developer has a contract they trust.
-
Execute — With the full contract loaded via CLI or MCP, the developer (or AI agent) implements with complete context. No Slack threads. No guessing.
Why developers adopt it
- Specs reference their actual codebase — not generic descriptions
- They have a voice BEFORE work begins — not after the PM "finalizes" the ticket
- The contract is clear — no "that's not what I meant" after three days of work
- AI agents execute better with Forge contracts — less rework, fewer corrections
Choosing Your Configuration
Start simple and evolve:
- Starting out? Use PM only to get familiar with ticket creation and the AEC format
- Ready for collaboration? Add developer review — even without a codebase connection, the review step dramatically improves spec quality
- GitHub connected? Enable codebase context for file-level precision in your specs
- AI agents in your workflow? Add
forge developfor the complete pipeline from spec to implementation
You can mix configurations across tickets. A quick bug fix might be PM-only. A major feature gets the full lifecycle. Forge adapts to each ticket's needs.