Geyser SDK · 0.3.4

Build apps that put Agents to work

Add Agents to the software your team or customers already use. They can look into a request, write up a brief, or suggest a change, then send back a result your application can work with.

“Tell me the job, the format you want back, and a limit to stay within. I’ll send back a result your app can use.”

Otis

Help support dig into a tricky ticket

Help someone on your support team make sense of a hard ticket and draft a reply. The support example sorts the ticket details you pass in, asks an Agent to look into the background, and then checks the sources the Agent cites.

Your app handles the help desk connection and passes in the records. The example writes a draft, and a person decides whether it’s right and whether to send it.

Try the support example

Turn research into a brief people can use

Build something that helps a sales team prep for a customer call, or helps a researcher get up to speed on a topic. Pass in the notes and documents and ask an Agent for a summary with references.

Start with the Agent review and source-checking examples. They catch references that are missing or made up, though you’ll still want your own review to check the facts.

Try the research building blocks

Add an approval step to an operations tool

Let an Agent propose a change to a record, show a person what will happen, and wait for their decision. It suits internal tools where staff review changes before they go through.

The approved-record examples run against a sample database or a throwaway Agent workspace, and they cover approval and recovery from an interrupted response. Any connection to a real business system comes from your app.

Try an approved update

Build tools around the way your team works

Make a proposal assistant or a project-handoff tool that follows your team’s instructions. Package up the steps and the background it needs, and you can reuse the approach on the next job.

An Agent Bundle keeps instructions, skills, and portable context together. Each run still follows the workspace’s permissions as they are at the time, and asks for a fresh approval before anything consequential.

Package a reusable workflow
Why build on Geyser?

Spend your time on the app while Geyser tracks the work

Your app owns the interface and its connections to other services. Geyser keeps track of the Agent’s task in one place inside the customer’s workspace, including its progress, budget, permissions, approval requests, and result.

If the connection drops, you reconnect to the same task. You can show people work that’s waiting on a decision, and get results back in the format your app expects. All of this pays off when a job takes several steps or needs oversight. For a simple text transformation, a direct model call or an ordinary function may be all you need.

Build an application with the SDK
  1. 1
    Give it a job

    Pass in information you’re allowed to use, the format you want back, and limits on time and spending.

  2. 2
    Start the task

    Reuse the same operation ID so your app can find the task again if the connection drops.

  3. 3
    Follow the work

    Read progress, and show people any request for approval or for help recovering an interrupted action.

  4. 4
    Use the result

    Fetch the saved result from the customer’s workspace, and check it before you use it.

Start with a small working example

Run your first tool on your own machine

Write a Python tool that takes JSON, run it on the sample inputs, and look at what comes out. The quickstart runs your real code in a restricted environment, along with a test that rejects invalid input, and you don’t need an account to try it.

The tagged-source quickstart needs Python 3.11 or newer and a supported sandbox, which is macOS sandbox-exec or Linux bubblewrap with user namespaces allowed. Once it’s installed, the handler doesn’t use a model, network access, or credentials.

Understand the extension contract
0.3.4 tagged release · local execution
git clone --branch v0.3.4 --depth 1 https://github.com/geyserlabs/geyser-open.git
cd geyser-open
python3 -m venv .venv
. .venv/bin/activate
python -m pip install -e ./sdk/python -e ./cli
geyser init tool word-count
geyser test word-count
geyser --json dev word-count
Authority and availability, stated plainly

Where things run, and what they cost

Each project works through one Agent you pick. The project’s keys work only for that project and only where its workspace’s data is currently kept. A signed JSON package goes live only after that Agent has checked it byte for byte and it has passed its tests in a sandbox.

A pure JSON extension can’t touch the computer’s files, the network, other processes, or any credentials it might inherit. Agent tasks you send through the SDK or CLI use the models and tools set up in your workspace. The SDK itself is MIT-licensed, and remote model and compute use is billed through your account. During the preview there’s no free compute allowance and no availability SLA.

SDK/CLI 0.3.4 is on PyPI and GitHub. Homebrew installs it on Apple silicon Macs and AMD64 Linux, and local Python handlers also need Python 3.11 or newer. Running locally works today. Running a task remotely needs workspace and Agent software that match, so check the Developers page in the console before you send one. A bundle doesn’t widen what an Agent may already do, and a model profile picks from models that are already set up. Public inspection forks don’t run tools, and replaying an execution is a separate operation.

Python SDK on PyPICLI on PyPIDownload SDK/CLI 0.3.4Read the changes and migration notes
Make a start

Start with one job your app could make easier

Pick an example, adapt it to how your team works, and see whether it saves you from building the task handling yourself. Each example is a starting point for your own application.