Managers and IT
Give your teams agents without losing control
Where the data lives, who can access what, what the agents do, what it costs and how to leave: the five questions your IT department asks, each with the screen that answers it.
Where is our data?
Wherever you choose to install it
The same software runs in three places: on our cloud, hosted in Canada or France, on your servers, or in an isolated environment with no Internet access.
- On your premises, one command installs everything: PostgreSQL for the database, Redis for the queue, S3-compatible storage for files.
- No usage telemetry. The only outbound call by default reads the model catalog, and it can be turned off.
- What you send to a commercial model goes to its provider; a local model (Ollama, vLLM) keeps everything on your side.
Who can access what?
One role per person and per space, checked before it is granted
The organization is split into spaces: a team, a department, a client. In each space, a person gets a role: Admin, Builder, Operator, Runner or Viewer, or a role you compose.
- A Runner runs agents without seeing their instructions; a Viewer looks without running anything.
- A guest, a contractor for instance, only enters the spaces you add them to.
- "Preview a role" shows the app through that person's eyes, before you grant the access.
What do the agents do?
Every run leaves a readable trail
A run is one launch of an agent. Its log keeps every message and every tool call with its arguments and result, and the "Execution" card says who launched it, how, and with which version.
- The trigger is named: a person, a schedule, an API key, a tool such as Claude Code.
- Admins and builders see every run in the space; other roles only see their own.
- Permission changes are recorded in the database. Viewing them in the app and through the API is on our roadmap.
What does it cost?
The cost of every run, measured while it runs
Every model call is counted. A run's "Usage" card gives its estimated cost, the model, the tokens and the context gauge. With your own model keys, the provider bills you directly.
- A run is capped at 30 minutes, 1,536 MiB of memory and 2 vCPUs.
- Per organization: at most 200 launches per minute and 50 concurrent runs.
- Totals per agent, period or team, and budgets, are on our roadmap. Until then, the runs API lets you add them up yourself.
What if we leave?
Your agents leave with you, in an open format
An agent exports as an AFPS file: its manifest, its instructions and its skills, described by an open specification. And when you self-host, the whole database stays yours.
- "Download bundle (with deps)" takes the agent and the skills it depends on, in a single .afps-bundle file.
- Skills follow the open Agent Skills format (a SKILL.md file): they can be read without Appstrate.
- The code is public on GitHub, under the Apache 2.0 license: your team can audit it.
What your IT checks
An honest status for each point: available, in progress, on our roadmap or not yet.
- Available
Hosting on your premises, down to an isolated environment
On your servers, with a local model and your images.
- Available
Data sent to models
Depends on the model you choose: a local model stays with you.
- Available
Encrypted secrets, never seen by agents
AES-256-GCM encryption, credential proxy.
- Available
Isolation of every run
One network and containers per run, with Docker.
- Available
Run history
Who, when, which model, what cost, which tool calls.
- On our roadmap
Viewable audit log
Permission changes are recorded in the database; viewing them in the interface is on our roadmap.
- Available
Team sign-in
Email and password, magic link, Google or GitHub.
- In progress
Data processing agreement and subprocessors
Being written.
- Coming soon
Enterprise SSO (SAML, Microsoft Entra)
With the Enterprise license, per organization.
- Not yet
SOC 2 and ISO 27001
Not started. The code is open for audit.
An open-source core, an Enterprise edition to come
The whole repository is Apache 2.0, with no usage limits. An Enterprise edition will add features bought by organizations, without taking anything away from the core.
Apache 2.0
The whole repository, which is the whole platform today.
- Agents, skills, integrations, Chat and Studio
- Roles, run history and costs
- Install on your servers or without Internet
- Fork, extend, host for others, even in a commercial product
Enterprise edition
Features bought by the organization rather than by those who build agents, in a separate ee/ directory.
- SSO configured per organization, SCIM provisioning
- Audit log search, export and retention
- Readable code in the same repository, license key checked offline
- Optional: the core works fully without it
Our commitments
What stays true, whatever the edition.
- A feature released under Apache 2.0 will never become paid
- No relicensing of the core or of contributions
- A community image without commercial code, always published
- If the Enterprise edition stops being maintained, it becomes open source
Sign in with Microsoft 365
Your teams will sign in with their work Microsoft account, with no new password to remember. Appstrate only accepts members of your tenant whose email domain is verified.
- No personal Microsoft accounts, no guests from another tenant.
- Existing sign-up rules still apply, such as the list of allowed domains.
- SSO configured per organization (SAML, Microsoft Entra) will follow, with the Enterprise license.
Frequently asked questions
How does the team sign in?
Can we restrict sign-up to our domain?
What happens when someone leaves the organization?
Do agents see our passwords and tokens?
Who pays for model calls?
Are you SOC 2 or ISO 27001 certified?
What is the status of the product?
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.
Let's talk about your case
Thirty minutes with us to scope a first use case and its gain.
Read the code
Apache 2.0. Your technical team can read it all, install it all, keep it all.