Dispatches
Build · Applications · Computers

Governed machines your agents can build on

A Computer is an agent-operable machine with files, code, database, browser, and web server—snapshotted after every message, promotable to production.

Athena Intelligence uses ten agents on Computer assets to run website, CRM, FP&A, sales ops, and infrastructure—roughly half the business.

Computer: sales-dashboardRegionQuarterModeSnapshotssavedBudget usedtrackedQ1Q2Q3Q4Owner: Analyst · owner-only
Fig. 01A Computer asset: file tree of workspace assets, terminal mid-run, browser preview, and snapshot timeline with one entry per message
  • Micro-VM isolation
  • Snapshot after every message
  • 24-hour undo window
  • Computer mode controls who may use terminals and files (owner-only, team, or workspace)
  • Promote to Application in one step
The problem

Agents need machines to do real work

An agent that can only talk cannot install a package, run a script, host a preview, or read a folder of 10,000 files. The moment real work starts, it needs a machine, and that machine needs the same governance as the person using it.

Most teams solve this with a laptop nobody audits or a cloud account nobody owns.

A computer for the agent. Governed like the rest of the platform.

Ungoverned accessGoverned ComputerSticky noteprod creds in Slack DMLaptop nobody auditsCloud account nobodyNo audit trail or ownershipOwner: AnalystMode: owner-onlysnapshotsBudget, mode, and snapshot count visible
Fig. 02Before: ungoverned credentials on sticky notes. After: Computer asset with owner, mode, budget, and snapshot count in the header
How it works

Boot. Work. Snapshot. Promote.

  1. Boot from an Environment. A template with startup commands (install packages, set variables), knowledge, resources, and a prewarmed sandbox. Your assets are mounted as a real folder tree the agent can traverse.

  2. Work. Code, data, a database, a browser, a web server, all in one place. The agent uses APIs where they exist and a governed, recorded browser where they do not.

  3. Snapshot. Automatically after every message. Roll back, or restore as a new computer. Twenty-four-hour undo window by default.

  4. Promote. When the app is right, promote the Computer to an Application asset: immutable versions, staging URL, custom domain, SSO-group access.

01Environmenttemplateinit · knowledge ·prewarmed02Computer assetfiles · code · DB ·browser · server03Snapshot permessageroll back ·restore-as-new04Promote toApplicationversions · domain ·SSOImmutable versions, staging URL, custom domain, SSO-group access
Fig. 03Boot from an Environment, work in a Computer with snapshots after every message, then promote to an Application with versions and SSO
People and agents

What people and agents each do

Apps. A full virtual machine per project with coding agents preinstalled. Describe an app, watch it get built, promote it to your people or your clients.

CapabilityPeople canAgents canTogether
AppsDescribe it, review the preview and the code, promote it.Write, test, and fix the code; ship new versions to staging.Nothing reaches users until a person promotes it. Rollback is one click.

Anything a person can do here, an agent can do with the same permissions, and both land in the same audit trail.

Use cases

What you can build on a Computer

A million-row CSV becomes a database.

Drop the file. Athena creates a Database asset with typed columns, builds indexes, and runs statistics before you ask. Analytical queries went from 30 seconds to about 500 milliseconds after tuning.

CSV droppedAnalystmillion-row fileSchema inferredAthenatyped columns ·indexes builtQuery runAnalyst0.48 sDatabase assetAthenaqueryable in secondsAnalytical queries went from 30 seconds to about 500 milliseconds after tuning
Fig. 04Drop a million-row CSV, Athena creates a Database asset with typed columns and indexes, query returns in 0.48 seconds

Our website runs on one.

athenaintel.com is built, hosted, and maintained by an agent on a Computer asset, then promoted as an Application. Copy, images, and code change by describing the change.

copy editsAthenadescribed changeappliedimage updatesAthenadescribed changeappliedcode changesAthenadescribed changeappliedv212Athenastaged 09:02,promoted 09:14athenaintel.com built and promoted from a Computer asset by Athena
Fig. 05The website version history: Athena builds, stages, and promotes by description

Legacy systems with no API.

A governed browser session with your credentials, recorded and attributable. Entra admin consoles, a legacy system via SFTP, a GovCloud tool where integrations are prohibited. Slower than an API call, and we never describe it as bypassing a control.

Entra admin consoleLegacy system via SFTPGovCloud toolGOVERNED BROWSER · RECORDED · ATTRIBUTABLEActing as: AnalystRecording pillactiveYour credentialsin sessionEvery actionloggedSlower than APIcallNever described asbypassing controlLegacy systems with no API operated through browser with full
Fig. 06Browser session with credentials, recording, and attribution for systems without APIs

Data-science tooling without a second vendor.

Notebooks with isolated compute and per-user volumes on billion-row data. Ephemeral Postgres (up to 1 TB), spun up and down on demand.

Data science on Computer assetsNotebooks · isolated compute · per-user volumes · billion rows—Ephemeral Postgres up—Spun up on demand—Spun down on demand—No second vendorCOMPAREVolume: AnalystAthena40 GB isolatedAthenaNotebook cellAnalystNotebook against lakehouse · isolated volume · ephemeral database
Fig. 07Notebook with isolated per-user volume and ephemeral Postgres up to 1 TB on demand

Agents debugging the cluster they run in.

Kubernetes monitoring and remediation inside a customer's own cluster. Pods, Kafka, S3 errors inspected; a pull request proposed. Right-sizing took one service from 46 GB to 3 to 5 GB.

Athena wants toPropose pull request for memoryright-sizing in customer clusterPods, Kafka, S3 errors inspectedService memory: 46 GB to 3–5 GBKubernetes monitoring and remediation inside clusterWAITING FOR A PERSONPull request proposed for review and mergeAgent debugging the cluster it runs in · recorded · attributable
Fig. 08Athena inspects cluster errors and proposes a PR that right-sized one service from 46 GB to 3–5 GB
What customers say
“Digital worker, it gets spun up, it gets given access to the toolkits that it needs. It gets given access to the data that it needs. Just exists for the period it needs to exist to perform the task. And then it gets torn back down again afterwards.”
Audit lead · A global professional services firm
“I finally got a chance to play around with the computer asset and build out the chatbot you mentioned. It was really impressive.”
Marketing lead · An AmLaw 100 firm
“So it's a full blown running Linux instance with all the bells and whistles that come with that.”
Technology lead · A global professional services firm
“It just feels better knowing I have a full copy locally”
Audit team · A global professional services firm
“It's good to understand the capability of the platform. That how far, and how quickly you guys can get these things going [...] that has been impressive overall.”
VP-level AI leader · A Fortune 500 retailer
“Lets us get a lot more out of our data than [...] We can sit on top of all of the data in [our data warehouse] as opposed to just, like, our little ingested, you know, three gigabytes semantic models. I think we have a lot of power here.”
Analytics lead · A Fortune 500 retailer
“Gave Athena the spreadsheet, explained which column I was trying to figure out, and then ten secs it told me... exactly how the calculation was based on the other data.”
Analyst · A global manufacturer
“He wrote the entire stuff. By itself. And I was there, and I was thinking, alright. If I had to write all this stuff, I mean, I would've spent days, days, days.”
Manufacturing analyst · A global manufacturer
“I send that to the programmer and he said, alright. I read your file. I copied the section of the code because he has the right interface, and it works. Wow. Just like magic.”
Manufacturing analyst · A global manufacturer
“One workspace. All different, let's say, avatars of their output data. Right? Suddenly, it's dashboard. Excel is here. PowerPoint is when it's all together. It's very impressive.”
Global Sustainability lead · A global manufacturer

Verbatim from customer calls. Customers anonymized.

Governance

Governance, stated plainly

  • Micro-VM isolation with warm pools; live resizing; shared storage.

  • Snapshots after every message; 24-hour undo; restore as a new computer.

  • Computer mode limits who may create computers, open terminals, or touch files; set to owner-only, team, or workspace.

  • Hard budgets per run. Approval gates on destructive actions.

What to plan for

A computer is a computer. Set a budget and an approval gate before the agent runs anything destructive. Both are one setting each.

FAQ

What is a Computer?

An isolated workspace where agents and builders create and test software safely.

Is it safe to let an agent build?

Yes. It is sandboxed, reaches only what you connect, and every action is logged.

How does it become something people use?

Promote it to an Application when it is ready.

One platform underneath

AGS · Athena Governance System

Records every change by a person or an agent, rolls back one contributor's edits without losing anyone else's, and keeps the model, instructions, and sources behind each agent action.

Palladium · deployment

How the platform is deployed: Athena's managed cloud, your cloud on AWS, GCP, or Azure, on-prem, air-gapped, or GovCloud. Same platform in every option.

How it fits together

Build, Work, and Data on top; 150+ connectors and every surface in and out; one map of the whole platform.

The rest of the platform

Related products and stories

Give the agent a machine this week.

Bring a folder of files nobody can query, or the app three people have been asking for. We will build it with you on a Computer and show you the undo button.