Platform overview
FYAI is a platform for building, running and governing AI agents. You give agents instructions, tools and knowledge. You combine them into teams and workflows, run them from chat, on a schedule or from external events, and keep a governed record of what they did.
This page is a map. Each section names the part of the app and links to the page that explains it.
How the pieces relate
Organisation
└── Project (most things live in a project)
├── Agents ── use ──► Models, Functions, Knowledge bases, Memory
│ └── grouped into Teams
├── Workflows ── run agents, teams and functions as steps
│ └── each run is a Mission
├── Programs ── multi-phase work whose activities run workflows, agents and teams
├── Functions ── use ──► Integrations (stored credentials) and Execution environments
├── Data sources ──► Dashboards (built from Elements)
└── Webhooks, Event sources, Schedules ── start workflows automatically
- Organisations and projects. Everything you build belongs to an organisation and, almost always, to one project inside it. Projects keep work separate (for example a team's sandbox and its production setup). You can move teams and workflows between projects with promotion. See Organisations and projects.
- Starter content. A new project is seeded with the platform's shared models, agents, teams and other resources unless you turn Include platform-managed resources off when creating it. Platform-managed items are read-only copies of FYAI templates.
The app, section by section
The left navigation is grouped the same way as these docs.
Dashboard and Chat
| Area | What it's for | Read |
|---|---|---|
| Dashboard | Your dashboards: tiles fed by live data sources, built by you or by an agent. | Dashboards |
| Chat | Talk to an agent or a team. Approvals, hand-offs and delegated work show up in the conversation. Chat is also available as a drawer on every page. | Chat, Chat drawer |
Build
| Area | What it's for | Read |
|---|---|---|
| Agents | An agent is a model plus instructions, tools (functions), knowledge and memory. | Agents |
| Teams | Several agents working together under one of six orchestration patterns. | Teams |
| Workflows | A visual graph of steps — agents, teams, functions, conditions, loops, approvals — that runs the same way every time. | Workflows |
| Missions | A single run of a workflow, with its inputs, live progress and results. | Missions |
| Programs | Long-running, multi-phase work. Each phase has activities carried out by workflows, agents, teams or people, with checkpoints for human sign-off. | Programs |
Resources
| Area | What it's for | Read |
|---|---|---|
| Integrations | Credentials and connection details for external services, stored as secrets. | Integrations |
| Functions | The tools agents call: your own code, MCP servers, integration operations, runner tools and more. | Functions |
| Models | Which LLMs your agents use — platform-provided models or your own keys — and model groups for routing. | Models |
| Knowledge | Knowledge bases of documents that agents search. | Knowledge bases, Choosing how to add knowledge |
| Data Sources | Live, refreshable data feeds for dashboards, agents and workflows. | Data sources |
Platform
| Area | What it's for | Read |
|---|---|---|
| Environments | Machines where agents' tools run — for example your own computer or server, connected with the FYAI runner. | Execution environments, Runner |
| MCP Gateway | Expose MCP tools to external agents and clients under governed identities, entitlements and containment. | MCP Gateway |
| Elements | The building blocks that dashboard tiles are rendered from. | Elements |
| Manifests | Structured documents that agents and other features read as configuration or reference data. | Manifests |
Governance
Risk classification, runtime approvals, AI system registration, policy decisions, alerts, a control library and an exportable assurance pack. Start at the Governance overview.
Marketplace and Settings
- Marketplace (sidebar footer) — browse, install and publish Stacks: installable bundles of agents, teams, workflows, dashboards and more. See Marketplace.
- Settings (avatar menu → Settings) — your account, projects, organisation, usage and billing, API keys and more. See Settings.
Ways work gets started
| Trigger | Starts | Read |
|---|---|---|
| A message in chat | An agent or team run | Chatting with agents |
| Launch Mission | A workflow run | Your first mission |
| A schedule | A workflow run | Scheduling workflows |
| An incoming webhook | A workflow run | Webhooks |
| An event source | A workflow run | Event sources |
| Running a program phase | The phase's activities | Running a program |
| The REST API or MCP server | Agents, workflows and most other actions | Developers |
| Microsoft Teams | A chat with an agent or team | Microsoft Teams bots |
Keeping humans in control
- Approvals pause a tool call or workflow step until someone approves it. Whether a call needs approval depends on its risk level, your project's risk appetite and any overrides on the team, agent or function. See Approvals and risk appetite and Using approval gates.
- Checkpoints hold a program phase until it is signed off. See Programs.
- Killswitches, history and observability let you stop an agent, roll back a configuration change and see what ran. See Observability.
Next steps
- New here? Follow Getting started.
- Looking up a term? See the Glossary.
- Building from code? Start with the Developer overview.