opencomputer

the infrastructure company of San Francisco

[Utpal Nadiger @utpalnadiger]

Have you checked out the new @opencomputerHQ UI?

Ask your agent to get itself a computer here: https://t.co/UXZ7dQNedg

[opencomputer @opencomputerHQ]

RT @utpalnadiger: Last night, we said screw it & built a playground for building a managed agent on @opencomputerHQ

https://t.co/0mNMdLNN…

[Igor Zalutski @IgorZIJ]

Today we shipped per-session models in Durable Agent Sessions on OpenComputer

You can now set a model for every session via API or SDK, and that will override agent's default model.

Grab your API key and try it! https://t.co/rDRdqXlxW9

[Igor Zalutski @IgorZIJ]

Today we shipped Schedules as part of Durable Agent Sessions on OpenComputer

You can now start your agent sessions automatically on a cron in addition to inbound Slack messages or API calls.

Grab your API key and try it! https://t.co/rDRdqXlxW9

[Utpal Nadiger @utpalnadiger]

Yes absolutely!

Build your own for your team with durable agent sessions on @opencomputerHQ

https://t.co/9RTt03ilEN

[opencomputer @opencomputerHQ]

Durable agent sessions in https://t.co/ZDMLeMmtKQ now has flue support!

[Igor Zalutski @IgorZIJ]

Today we shipped support of Flue in Durable Agent Sessions on OpenComputer

> oc agent create myagent --runtime flue
> oc agent deploy
> oc session create --input "hi Flue"

Try it! Grab your API key here: https://t.co/rDRdqXlxW9

ps very curious what @FredKSchott thinks of this

[Utpal Nadiger @utpalnadiger]

Some big updates this week! The https://t.co/Yz61qJLcD0 adapter went live on https://t.co/3qLf8CcdxD by @colemurray.

We spoke to him a few days back and filmed Ep 1 of the picks and shovels podcast.

Had a lot of fun chatting with him about background agents, sandboxes and his hot takes on the space. Give it a listen! (sorry about any editing/subtitle errors, we're learning!)

00:06 Cole Murray Intro & Background

01:01 The idea behind open-inspect

02:00 Issues with off the shelf products


03:19 November 2025 model releases causing a step change


03:46 Ramp’s implementation and Cole’s live tweets


04:32 Usecases for background agents in the enterprise

06:48 Have background agents broken containment?

08:11 Is background agent usage determined by org size?


09:35 Harness in Sandbox vs Harness outside sandbox


10:55 Where you put the secrets matters


12:10 What are the features the perfect “agent sandbox” should have?


14:56 Memory snapshots are actually not that important?


16:15 Off the shelf vs Building your own background agent


17:20 Why owning the architecture kinda makes sense


18:48 Cole’s hot take on background agent businesses


21:03 sandbox providers will have their own managed agents systems


22:00 The managed agents conversation


23:40 Cole’s AEO maxxing skills

[Igor Zalutski @IgorZIJ]

Today we shipped support of pi runtime in OpenComputer Durable Agent Sessions API

pi is _awesome_ - i didn't quite realise just how awesome it is until building this. It also allowed us to improve the runtime contract. custom runtimes coming soon! sneak peek in the video

very curious what @badlogicgames thinks of the approach we are taking with Durable Agent Sessions

[Utpal Nadiger @utpalnadiger]

Really enjoyed reading this and also resonated with this bit specifically. Its something we hear from a lot of our users too!

To build your own "junior" or say your teams own "claude tag" in slack (like @tryramp's inspect or @coinbase's Mux) means jumping through a million hoops and braving an ungodly amount of papercuts.

We're trying to make this as simple asy possible and are working really hard to make it easy.

We're calling our implementation the durable agent sessions API. Give it a shot and tell us what sucks?

https://t.co/lYhk2hbGCI

[Igor Zalutski @IgorZIJ]

Today we shipped Watches as part of Durable Agent Sessions API on OpenComputer

Thansk @zeeg for the idea!

Watches allow your agents to selectively subscribe to GitHub events - but only from those PRs that they have access rights to

Grab your API key and try it: https://t.co/kWkWzSvsLb

[Igor Zalutski @IgorZIJ]

Today we shipped Revisions for agents running on OpenComputer Durable Agent Sessions

You can now roll back to any version of prompts and skills; deploy from CLI or dashboard; or connect a GitHub repository to auto-update your agents' skills and prompts from it

An agent is just a directory!

Grab your api key here: https://t.co/kWkWzSvsLb

[Utpal Nadiger @utpalnadiger]

giving agents computers today at AIE in SF

come chat with the https://t.co/qPwpgFRfs4 team if you’re around!

[Utpal Nadiger @utpalnadiger]

Stupidly simple way to build your own claude tag for your team - now live on https://t.co/Yz61qJLcD0 via durable agent sessions.

We kept seeing mutiple users build background agents inside slack, and it also quickly became one of our fastest growing use cases internally.

Naturally, we are trying to make this as easy as possible.

Go give it a shot!

[Igor Zalutski @IgorZIJ]

Today we shipped built-in model credentials in OpenComputer Durable Sessions

Having model keys stored securely is nice but what if you just want to give it a try? Well, now you can - new agents by default are assigned a built-in credential, billed to your OpenComputer credits via OpenRouter

We currently pass OpenRouter billing as-is without markup. This may change in the future.

Give it a try! https://t.co/kWkWzSvsLb

[Utpal Nadiger @utpalnadiger]

Every enterprise we talk to is building their own model agnostic internal agents (stuff like Claude Tag but homegrown and multi channel - @tryramp has inspect, @WorkOS has horizon, @coinbase has mux).

It's table stakes now.

Today, we shipped a simple way to build a claude tag alternative for your org. Anyone in the team (think sales, customer support, gtm, pm's etc) can @mention it in a channel to start a session and steer it from there.

It runs on opencomputer durable agent sessions (https://t.co/zEN7tiZRSg).

We'd love any and all feedback. We're very excited for this!

[Igor Zalutski @IgorZIJ]

Today we shipped Slack support in Durable Agent Sessions

You can give every agent you build on OpenComputer its own Slack handle

Now you can create an agent like Claude Tag or Devin without writing any code! API and SDK also work ofc

Try on https://t.co/gueuPyqAUW

[Utpal Nadiger @utpalnadiger]

every single one of these agents will need a real computer

give your agent one or even better, ask it to get one for itself

https://t.co/qPwpgFRfs4

[Igor Zalutski @IgorZIJ]

Today we shipped Credentials in Durable Agent Sessions, handy for storing model keys

Two levels of security:
- Dedicated secret vault at rest (@infisical)
- Synthetic key + egress proxy swap in flight

A rogue agent physically cannot read your model keys

[opencomputer @opencomputerHQ]

WE LOVE TO SEE THIS!

An email triage agent to address bogus security reports built internally at @xata.

Uses https://t.co/ZDMLeMmtKQ as the computer for the agent and @resend for email infra!

[Tudor Golubenco @tudor_g]

In a case of fighting AI slop with AI slop, I created an agent to triage the tons of (mostly bogus) security reports we receive on our bounty program.

The agent runs in an https://t.co/ZPFsWrBQZD sandbox and has access to our source code, so it can validate or refute most claims with high accuracy by checking the source code.

Because we receive these reports by email, I have the agent connect directly to my inbox, check entries labeled with a particular tag, and email me the verdict and a sample response that I can then c&p.

I always read the report to check, but most of the time I can just send the suggested response as is. If the original report sounds like AI slop, I don't feel bad for sending AI slop back.

[Igor Zalutski @IgorZIJ]

Durable Agent Sessions now available in OpenComputer UI - which also got a refresh

You can create agents, see all their sessions and sandboxes they are working in, register webhooks

Grab your API key and give it a try!