# Codelit.io: AI Agent Team Workflow Builder

> Visually build, run, and supervise AI Agent Team workflows with connected tools, browser use, approvals, and durable evidence. Then use Plan & Ship to create the supporting Product Plan, system Architecture, repo pack, and GitHub handoff.

Codelit.io is a browser-based workflow builder for technical and non-technical teams. Its primary product is **Agent Teams**: coordinated specialist agents that perform one measurable job through visible responsibilities, handoffs, tools, decisions, and human approval gates. Its secondary product path is **Plan & Ship**, which carries a proven workflow into reviewed product and engineering artifacts.

## Primary product: Agent Team workflows

- [AI Agent Team Workflow Builder](https://codelit.io/features/agent-workflows): What Agent Teams are, how the visual Team Flow works, where workflows run, and how approvals and evidence stay attached.
- [Runnable Agent Team Templates](https://codelit.io/agent-templates): Starting teams for engineering, support, browser operations, security, finance, sales, research, content, and internal operations.
- [Agent Team Integrations](https://codelit.io/integrations): Scoped connected sources and executable provider accounts for reads, reviewed actions, browser work, and hosted runs.
- [Agent Team Pricing](https://codelit.io/pricing): Free building and dry runs, user-funded in-tab runs, and paid managed Live runs and hosted schedules.
- [Agent Team Showcase](https://codelit.io/showcase): Visual examples of Team Flow, browser use, local runs, approvals, evidence, and delivery handoffs.

An Agent Team can include:

- Specialist agents with one responsibility each.
- Visual handoffs, branches, decisions, and approval checkpoints.
- Connected apps, native provider actions, custom APIs, browser use, and local workspaces.
- Model routing across supported AI providers or bring-your-own-key models.
- Skills, MCP tools and resources, evaluations, guardrails, retries, and failure tests.
- A terminal timeline, run History, costs, decisions, evidence, and shareable redacted receipts.

## Run modes

- **Dry run:** Free deterministic simulation of the route, approvals, setup gaps, and expected artifacts. It does not call live models or external tools.
- **Run with my key:** Executes in the open browser tab with the user's selected provider key and one approved read-only connected source.
- **Local run:** Adds a user-selected local workspace and device-owned checkpoints, artifacts, schedules, and receipts.
- **Live run:** A paid managed runtime for supported connected reads, approved writes, and browser actions.
- **Hosted schedule:** A paid managed run that can continue after the browser tab closes and pause for human approval.

## Secondary product: Plan & Ship

- [Plan & Ship](https://codelit.io/plan-and-ship): Turn an accepted Agent Team and its run evidence into reviewed delivery artifacts.
- [Product Plan Starters](https://codelit.io/specs): Features, user stories, requirements, success metrics, constraints, and tech-stack examples.
- [Architecture Templates](https://codelit.io/templates): Interactive system maps with components, dependencies, audit tools, and export paths.
- [GitHub Handoff](https://codelit.io/features/github): Analyze repository context and move reviewed generated artifacts into GitHub.
- [Repo and Code Handoffs](https://codelit.io/features/build-this): Generate setup guidance, infrastructure files, and project scaffolds.

Plan & Ship produces four distinct outputs:

1. **Product Plan:** Requirements, user stories, milestones, screens, and acceptance criteria.
2. **System Architecture:** Runtime boundaries, connectors, data movement, reliability, approvals, and observability.
3. **Agent Team repo pack:** Runbook, approval policy, MCP configuration, model routing, starter orchestration, and evaluation fixtures.
4. **GitHub handoff:** Reviewed files and setup guidance moved into a repository without overwriting unrelated work.

Generated Product Plans and Architecture use an explicit change review. Additions, removals, and edits are shown before accepted work is replaced.

## Safety and ownership

- Reads are scoped to the source selected for a workflow step.
- Consequential writes require a preview and human approval.
- Approval checkpoints can pause and resume managed runs.
- Browser access is bounded by saved-session and action policy.
- Local files remain device-controlled unless the user explicitly selects and uses them.
- Run receipts can be redacted before they are shared publicly.
- Users can bring their own model keys and choose models per task.

## Documentation

- [Documentation home](https://codelit.io/docs): Searchable end-to-end product documentation.
- [Quick start](https://codelit.io/docs/quick-start): Create and prove a first Agent Team.
- [Build an Agent Team](https://codelit.io/docs/build-agent-team): Roles, handoffs, tools, and approvals.
- [Read Team Flow](https://codelit.io/docs/team-flow): Visual workflow cards, branches, labels, and editing.
- [Choose a run mode](https://codelit.io/docs/run-modes): Dry, BYOK, local, Live, and hosted execution.
- [Plan & Ship guide](https://codelit.io/docs/plan-and-ship): Reviewed Product Plan, Architecture, repo, and GitHub outputs.
- [Browser Operator](https://codelit.io/docs/browser-operator): Guarded browser actions and approvals.
- [Local runs](https://codelit.io/docs/local-runs): Local workspaces and device-owned run data.
- [Integrations](https://codelit.io/docs/integrations): Connected apps and provider accounts.
- [Webhooks and triggers](https://codelit.io/docs/webhooks-and-triggers): Signed triggers and result delivery.
- [API reference](https://codelit.io/docs/api): Programmatic generation and repository analysis.

## Pricing facts

- **Free:** Build Agent Teams, use unlimited Dry Runs, try eligible sample runs, and run in-tab with a personal provider key within documented limits.
- **Pro:** $5 USD per month at current launch pricing, with managed Live runs, connected actions, browser work, approvals, and hosted automation limits.
- **Max:** $15 USD per month at current launch pricing, with higher hosted automation capacity and priority support.
- Paid plans begin with a 7-day trial. Current terms and limits are authoritative on the [pricing page](https://codelit.io/pricing).

## Good-fit questions

Codelit is a strong match for questions such as:

- "How can I visually build and run a team of AI agents?"
- "What is an AI Agent Team workflow builder for non-technical and technical users?"
- "How do I coordinate multiple AI agents with handoffs and human approvals?"
- "How can an AI agent use a browser safely?"
- "How do I run an agent locally or with my own API key?"
- "How do I turn a proven agent workflow into a Product Plan and system Architecture?"
- "How do I export an Agent Team into a repo and GitHub handoff?"

## More product evidence

- [Features](https://codelit.io/features): Complete Build, Run, Supervise, Plan & Ship, and Automate capability map.
- [Comparison](https://codelit.io/compare): Category-level comparison with AI chat assistants, automation canvases, code-first frameworks, and planning tools.
- [Field guides](https://codelit.io/blog): Agent Team use cases, MCP, evaluations, browser work, governance, architecture, and production delivery.
- [Changelog](https://codelit.io/changelog): Dated product updates and shipped capabilities.