Agents

One agent per task, built once for the whole team

An agent is an AI model working in a loop with tools, inside a sandbox. You give it a goal and instructions, it picks the steps.

What an agent is made of

What you give it

  • Instructionswhat it must do, in plain words
  • Skillsthe team's methods
  • Your connected toolsGmail, HubSpot, QuickBooks… through integrations
  • An AI modelClaude, GPT, Mistral or a local model
AgentIn a loop, in its sandbox

What it returns

  • A structured resultfields ready to reuse
  • Filesreports, drafts, exports
  • The run logevery step, every tool called

Create

Three ways to create an agent

By hand with a short form, by importing a package, or by describing it to the chat. The rest is set afterwards in its Settings.

  • "Create manually": a name, what it does and its instructions are enough.
  • "Import a package": .afps, .afps-bundle, .zip or GitHub repository.
  • "Build with the Chat": describe the task, the assistant writes the agent.

Versions

A draft to work on, versions to run

You edit the draft without disturbing anyone. Launches use the latest published version, which never changes.

  • "Create a version" freezes the draft: patch, minor or major.
  • "Compare" and "Restore" from the version history.
  • An agent exports to the open AFPS format, to be reviewed or reused elsewhere.

Sharing

Created in one space, used in others

The home space keeps control of the agent. You share it with another space or a person, who chooses to enable it.

  • Only those who write in the home space edit, publish or delete the agent.
  • A space that receives the agent decides to enable it; disabling keeps its settings.
  • Withdrawing a share also disables the agent's schedules in that space.

What it changes

For your teams

  • Run a ready-made agent, with nothing to configure.
  • Improve the draft without breaking what your colleagues use.
  • Find every run of the agent in its Runs tab.

For the organization

  • The same task done the same way across the team.
  • Frozen versions: you know what ran, you can roll back.
  • A home space that controls changes, spaces that choose what they enable.

Example agents to build

These are not agents shipped ready-made: they are ideas your teams build themselves, in a few minutes, with the existing integrations. They run on a schedule or on demand.

  • Example

    Lead qualification

    Trigger
    Every morning, 8 am
    Result
    Records updated and reply drafts
    Try it
  • Example

    Quote follow-up

    Trigger
    On demand
    Result
    Follow-up drafts in Outlook
    Try it
  • Example

    Meeting preparation

    Trigger
    Every day, 5 pm
    Result
    Summary sheet for tomorrow's meetings
    Try it

Frequently asked questions

Do I need to code?
Not to use the chat and Studio: you describe the agent, pick its tools and model. Code stays possible, with the API, the CLI and packages, for those who want it.
Can an agent act on my behalf?
Only with the tools and permissions you give it. With read permissions, it cannot send or change anything.
How do I run an agent from my code?
Through the API (an API key bound to a space), the command line or the MCP server. Each instance describes its API in OpenAPI. There is no official SDK yet.

Let's see what it would look like on your side.

Try it on your own in a few minutes, or take thirty minutes with us to scope a first use case.

To try it

Try it for free

Create a cloud account and describe your first agent.

For a guided start

Let's talk about your case

Thirty minutes with us to scope a first use case and its gain.

To audit

Read the code

Apache 2.0. Your technical team can read it all, install it all, keep it all.