Blaxel (@blaxelAI) on X (original) (raw)
Blaxel keeps infinite, secure sandboxes on automatic standby to run AI code, while co-hosting your agents for near instant latency.
- Pinned

What if Lambda and EC2 had a baby? AI Agents are a new breed of workload. They need the instant startup of serverless, but the persistence of a full server for the context. Traditional clouds force you to pick one, we decided to build both. We moved to bare metal to give 
Thrilled to announce that Blaxel has raised a $7.3M seed round led by
@firstround
, with
@ycombinator
,
@Liquid2V
,
@transposevc
, and incredible angels! π We're on a mission to build the foundational infrastructure for the next generation of software: AI Agents.
Ready to welcome the new
@ycombinator
S25 batch with iced coffee! βοΈ
We just launched sandboxes made for agents They are uVMs optimized for coding agents β meaning MCP-native and super low cold starts (boots in ~20ms and hibernates after 1 second of inactivity).
New: context from ANY documentation, ANY repository for your codegen agent with Nia by
@nozomioai
. One line SDK integration. Now available as a remote MCP on Blaxel.
Okibi is built on Blaxel Sandboxes! With 25ms boot times & persistent memory state, sandboxes provide secure dev environments to preview agents created by Okibi in real-time. Congrats
@MahyadGhassemi
and
@thesauravmitra
!
Launching our integration with
@agentmail
to give your agents their own inbox. Here weβre talking with a customer support agent that can access any doc in the world in real-time β all via emails.
New: Set time-to-live (TTL) for your Blaxel sandboxes, to automatically clean up after experiments. Perfect for short-lived environments where your agents can run adhoc scripts without requiring state persistence.
Replying to @blaxelAI
During YC, we had the privilege of working with amazing agent-native companies like
@humanbehaviorai
,
@vybe_build
&
@jazzberryai
. We're already processing millions of agent requests daily across 16 global regions, saving our partners up to 50% on their infra costs.
how to auto-detect bugs from screenshots in your Linear tickets we built a Linear knowledge base using
@airweave_ai
, and have an agent search it through our MCP server integration
Replying to @blaxelAI
And of course we vibecoded a codegen agent that uses sandboxes itself Here is a chat-based Next app of a code generation agent built with AI SDK that previews the code in real-time inside a remote Blaxel sandbox: github.com/blaxel-ai/sdk-β¦
Virtual machines for agents are inbound...!
Replying to @blaxelAI
A huge thank you to our investors for believing in our vision and to our customers for trusting us with their agents. The future is autonomous. Your infrastructure should be too. π
We brought MCP to serverless workloads, using
@supercorpai
Supergateway. This lets us run stdio-based MCP servers over WebSockets, which we found more reliable for serverless executions. Check out our blog post for a technical deep-dive ->