Sanaf AI
Start free
AI Document Extraction

VLM Run, worked by an AI employee

Not another app for your team to learn, and not a set of rules for you to build. You hire an AI employee, you give it access to your VLM Run account, and you ask it for things in Slack or Teams the way you would ask anyone else on the team. There is nothing to map and nothing to maintain.

  • AI Document Extraction
  • Connects with a key
  • 11 ready-made tools
  • Asked in Slack or Teams
At work

An illustration of how the conversation reads, not a transcript from a client account.

What VLM Run is

VLM Run provides multimodal agents, structured extraction, predictions, files, skills, feedback, and evaluation APIs.

You paste in a key from your own VLM Run account. It is kept encrypted, and you can withdraw it from inside VLM Run whenever you like.

How a run works

One message in, the work done in VLM Run.

  1. 01

    You ask

    You message your AI employee in the chat your team already has open, in one sentence, in your own words.

  2. 02

    It works out the step

    It decides what the job needs in VLM Run. You do not pick anything from a menu or wire anything together.

  3. 03

    It does the work

    Signed in to your own VLM Run account, inside the access you granted it, doing the thing you asked for.

  4. 04

    It reports back

    It tells you what it did, in the same thread. If it could not do something, it says so rather than guessing.

11 tools it already has

What it can do in VLM Run

These are the ready-made tools your AI employee already has in VLM Run. It is not limited to them, but it never has to be taught these.

  • Create SkillCreate a reusable skill from exactly one uploaded zip, prompt, or chat session.
  • Discover Extraction SchemasList supported structured-extraction domains, or return the full JSON schema for one domain when domain is provided.
  • Execute AgentStart a VLM Run agent execution from an existing agent name or inline configuration over multimodal inputs. Execution may consume credits and is asynchronous by default; poll the r
  • Extract Structured JSONStart structured JSON extraction from images, a document, a video, or audio using a domain, custom schema, or skill. Extraction may consume credits; document, video, and audio runs
  • Find FilesList uploaded files or find one by file ID or MD5 hash. In list mode, use offset and limit until hasmore is false.
  • Find SkillsList VLM Run skills or find one exact skill by ID, name, and optional version. In list mode, continue from nextoffset while hasmore is true.
  • Get RunGet the current status and result of one structured-extraction prediction or agent execution; call repeatedly to poll asynchronous work.
  • List AgentsReturn agents available to the connected account for selection before execution.
  • List ArtifactsList artifact metadata belonging to exactly one chat session or agent execution. Use offset and limit to traverse pages until hasmore is false.
  • List RunsList structured-extraction predictions or agent executions for the connected account. Use offset and limit to traverse pages until hasmore is false.
  • Upload FileUpload a local file to VLM Run for extraction, agent input, or skill creation. Retain the returned file ID for tools that consume uploaded files.
Access, and who is in charge of it

You stay the boss of your VLM Run account.

Nothing connects until you approve it. You grant access one app at a time, you can take it back the same way, and everything your AI employee does in there is written down where you can see it.

How we handle your data
  • It signs in to your own VLM Run account. You are not moving anything into ours.
  • You grant access one app at a time, and you can take it back the same way.
  • Everything it does in VLM Run is written down, with what it did and when.
  • Anything you tell it to check with you first, it checks with you first.
Other apps in the same corner of the business

It works in VLM Run and the rest of your software in the same job

A real job rarely stays in one place. Reading a message in one app, checking a record in another and writing the result in a third is one request to your AI employee, not three.

See every app

Questions people ask about VLM Run

Can an AI employee really work inside VLM Run?
Yes. It signs in to your own VLM Run account and works in it the way a new hire would, from the chat your team already has open. Nobody installs anything, and nobody learns a new screen.
Do I have to move anything out of VLM Run?
No. Nothing moves and nothing is replaced. Your records stay in VLM Run, your team carries on in the same screens they used yesterday, and your AI employee works alongside them in there.
How does it get into my VLM Run account?
You paste in a key from your own VLM Run account. It is kept encrypted, and you can withdraw it from inside VLM Run whenever you like. You approve it before anything connects, and you can take the access back the same way you gave it.
What can it actually do in VLM Run?
It has 11 ready-made tools in VLM Run today, among them Create Skill, Discover Extraction Schemas and Execute Agent. The full list is on this page. It is not limited to those, but it never has to be taught them.
Can it use VLM Run and the rest of my software in the same job?
Yes, and that is usually the point. Reading a message in one app, checking a record in VLM Run and writing the result somewhere else is one request to your AI employee, not three separate ones you stitch together.
Who decides what it is allowed to do in VLM Run?
You do. You grant access one app at a time and can withdraw it at any time, anything you ask it to check with you first it checks with you first, and everything it does in VLM Run is written down with what it did and when.