How it works
A delivery team of agents, in a workspace you control.
Portfall combines three things teams usually stitch together themselves: a cloud development environment per developer, a coordinated set of AI agents, and a governed path to any model.
Run · Add retry budget to webhook delivery
acme/payments-api · Standard lane · autonomy A3
- Orchestrator
Picked up PROJ-482 from the tracker
open-weighteffort: low - Explorer ×3
Mapped the affected modules (read-only)
smalleffort: none - Planner
Wrote the plan, 3 slices; approved
frontiereffort: high - Builder
Slice 2 of 3: mise run verify passing
open-weighteffort: medium - Reviewer
Independent review on a different model family
frontiereffort: high
Illustrative. Each role runs as its own isolated session with its own model and reasoning effort.
Agents
Specialist roles. Each one isolated, each with its own model.
You talk to the Orchestrator. It delegates to specialists, and every one of them runs as a separate session with its own model, reasoning effort and permissions. Delegation is visible as a tree you can follow and steer.
| Role | What it does | Typical model | Can it change code? |
|---|---|---|---|
| Orchestrator | Talks with you, frames the work, delegates, tracks progress, merges when policy allows | Strong open-weight | No; it merges approved work |
| Planner | Turns intent into a reviewed plan of small, testable slices | Frontier, high effort | Plans only |
| Builder | Implements one slice at a time and runs your verify command until green | Open-weight coder, escalating when stuck | Yes |
| Reviewer | Independent review of the diff against the plan, on a different model family | Frontier, high effort | No; review and approve only |
| Explorer | Fast, read-only research across code and docs | Small, fast model | No |
| Tester | Writes and runs tests, and drives previews in a browser | Open-weight coder | Tests only |
| Investigator | Investigates bugs and incidents in your logs, metrics and data, read-only | Strong open-weight | No |
Workspaces
Where agents and developers work together.
Every developer gets a persistent cloud workspace with an in-browser IDE, a terminal and authenticated preview URLs for any port. Start a dev server and open it on your phone; the agent can open it too, in its browser.
- Persistent, not ephemeral
- Your checkout, caches and tools survive between sessions. Workspaces sleep when idle and wake in seconds.
- Same toolchain for humans and agents
- Toolchains and tasks come from your repo’s mise configuration, so everyone runs identical versions and one `verify` command.
- Preview URLs
- Any port, behind single sign-on, shareable within your org when you choose.
Models
Any model. Your account or ours.
Every role can use any model available through OpenRouter, from frontier models to low-cost open-weight models. Pick a preset or tune each role’s model and reasoning effort yourself.
- Bring your own OpenRouter account
- Connect your organization’s account and we mint short-lived, spend-limited keys per workspace, visible in your OpenRouter dashboard.
- Or use ours (later)
- Managed model access with budgets, alerts and per-role cost reporting, billed through your plan. Coming after launch; bring your own keys until then.
- Governed routing
- Model and provider allowlists, zero-data-retention routing and region controls, enforced for every request.
Integrations
Fits the tools you already run.
Planned: GitHub, GitLab, Forgejo and Gitea first, then Bitbucket and Azure DevOps, for code; Jira, Linear and Notion for planning; Slack for approvals; and a governed gateway for any MCP server.
Deployment
Our cloud or yours.
Run on our multi-tenant cloud in the US or EU or in a dedicated environment, let us operate it in your own cloud project, or self-host a licensed edition on your own Kubernetes, including fully air-gapped with open-weight models on your own hardware.