Blaxel
Also known as: blaxel.ai, Blaxel Sandboxes
Blaxel, owned by Baseten since September 2026, runs agent sandboxes as microVMs that resume from standby in about 25ms, with agent and MCP server hosting, batch jobs, a model gateway and a secret injecting egress proxy.
Blaxel sells infrastructure for autonomous agents, covering sandboxes, hosting for agents and MCP servers, batch jobs, shared storage and networking, all driven from one API. Baseten acquired Blaxel in September 2026, and Blaxel continues to sell the platform under its own name while Baseten plans to build new products on its primitives, starting with sandboxes. Teams running on it include Webflow, Shortwave, Framer, Stripe, Airbyte and Sapiom, which ran 2.5 million sandboxes in three months.
Each sandbox is a microVM that boots in milliseconds, suspends itself after about five seconds without activity and resumes from standby in about 25ms with its memory and running processes intact, for as long as months later. Sandboxes can be snapshotted, rolled back, forked and archived, run on a cron schedule, expose preview URLs, and carry a built in MCP server that lets an agent operate the files, processes and ports.
Agent Drive is a shared filesystem that several sandboxes or agents mount at once, and volumes keep data for years. A managed egress proxy injects secrets into outbound requests and filters domains, methods and paths, and the model gateway routes calls to OpenAI, Anthropic, Mistral, Cohere, xAI, DeepSeek, Azure AI Foundry or Hugging Face models through one endpoint.
Pricing is pay as you go with up to $200 in free credits. Active sandbox compute costs $0.0000115 per GB of RAM per second, with no compute charge on standby, and standby snapshots cost $0.20 per GB a month. Blaxel holds SOC 2 and ISO 27001 and signs HIPAA BAAs as a $250 a month add on, with admin and member roles, SAML single sign on and SCIM provisioning. Sandboxes run in Oregon, North Virginia, London or Frankfurt, location policies keep workloads inside allowed regions, and a bring your own metal option runs the sandbox runtime on servers in the customer's own VPC.
Blaxel fits teams whose agents sit idle between bursts of work and must come back instantly with their state, and teams that want sandboxes, MCP hosting and model routing from one vendor. It fits less well for buyers who need browser control, approval steps or evaluation built in, or who want a vendor whose roadmap is not tied to a new owner.
Vendor details
Canonical URL
https://blaxel.ai
Category
Agent infrastructure
Funding status
Blaxel, Inc. was acquired by Baseten in September 2026 and still sells Blaxel under its own name. It had raised a $7.3M seed round led by First Round in December 2025.
Company status
acquired
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
TypeScript, Python and Go SDKs, a REST API, the bl CLI and an MCP server manage Blaxel resources. Sandboxes integrate with Claude Code, Codex, Amp, OpenCode, OpenClaw, the Claude Agent SDK, Claude Managed Agents, the OpenAI Agents SDK and Agents API, Mastra, Herdr, Omnara, DeepSeek Harness and NVIDIA OpenShell, and hosted agents run LangChain, CrewAI, LlamaIndex, Google ADK, PydanticAI and the AI SDK. The MCP Hub offers prebuilt MCP servers such as Gmail and Google Maps, and the model gateway connects OpenAI, Anthropic, Mistral, Cohere, xAI, DeepSeek, Azure AI Foundry and Hugging Face.
In practice
Your agent waits minutes or hours between steps, and you don't want to pay for idle machines. A Blaxel sandbox suspends itself when idle and resumes in about 25ms with its memory and running processes as they were.
Several agents have to work on the same files at once. Mount one Agent Drive into each agent's sandbox and every agent reads and writes the same shared filesystem in real time.
Your agent calls a paid API, but the key must never sit inside the sandbox. Blaxel's egress proxy injects the secret into outbound requests to that domain and blocks domains you haven't allowed.
Sources & related URLs
Research sources
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents reach outside systems through hosted MCP servers. The MCP Hub offers prebuilt servers such as Gmail and Google Maps, customers deploy their own MCP servers as autoscaling endpoints on the same private network as their agents, an OAuth flow lets end users authorize integration connections, and code mode turns any OpenAPI specification into an MCP server whose tools the agent calls by running TypeScript in a sandbox. Outbound calls from a sandbox can also carry secrets injected by the managed egress proxy, so the agent's code never holds the credential. Sourcedocs.blaxel.ai/functions/overviewread 2026-11-05 |
|
| Workflow Orchestration | Partial |
|
Batch jobs fan work out across many isolated sandboxes, running tasks in parallel with configurable concurrency, timeouts and retries, and executions can be started, monitored, retried and cancelled from the SDKs. Agents built with LangChain, CrewAI or other frameworks deploy as hosted endpoints, and n8n can orchestrate them. How an agent's own steps are sequenced and handed off stays in the customer's framework, and Blaxel's own session runtime for long running agents is announced but not yet available. Sourcedocs.blaxel.ai/jobs/overviewread 2026-11-05 |
|
| Knowledge Grounding & RAG | Not documented |
|
No retrieval index or knowledge base over the customer's content comes with the platform. Agent Drive, volumes and the sandbox filesystem hold files an agent's code reads, and retrieval built with LlamaIndex or another framework runs as the customer's own hosted agent. Sourcedocs.blaxel.ai/agent-drive/overviewread 2026-11-05 |
|
| Human Oversight & Guardrails | Not documented |
|
There is no approval step or escalation to a person for an agent's actions. Domain, method and path filters on the egress proxy limit what code can reach, and guardrails come from a partner integration with Rippletide, but nothing pauses an action until someone reviews it. Sourcedocs.blaxel.ai/sandboxes/proxy-domainsread 2026-11-05 |
|
| Security, Identity & Governance | Full |
|
Blaxel holds SOC 2 and ISO 27001, with reports in its compliance portal, and signs HIPAA Business Associate Agreements as a monthly add on. Workspace admins create policies, invite and remove users, change permissions and manage integrations, while members work with agents, MCP servers, model APIs, jobs and sandboxes and can only read policies. SAML single sign on and SCIM directory sync are configured at the account level, service accounts give machines their own credentials, and domain capture controls how a company's users sign in. Sourcedocs.blaxel.ai/security/workspace-access-controlread 2026-11-05 |
|
| Observability & Auditability | Full |
|
Agents hosted on Blaxel are traced with no setup. Each trace breaks a request into spans for LLM calls, tool calls and agent steps with their inputs and outputs, a sampled 10% of executions are kept and any run can be forced to keep its trace, and custom spans add detail. Logs cover request status, builds and custom output, metrics show request rates, latency percentiles and token use, and the managed egress proxy audits every request an agent sends out. Sourcedocs.blaxel.ai/observability/overviewread 2026-11-05 |
|
| Memory & State Persistence | Partial |
|
A sandbox on standby keeps its memory and running processes and resumes in about 25ms, even months later, and snapshots, forks and archives copy or preserve that state. Volumes keep data for years, and Agent Drive, free while in beta, gives several agents one shared filesystem for files and session context. This is machine and file state the developer manages, not a memory layer with a stated scope and lifetime that a buyer can review or delete as memory. Sourcedocs.blaxel.ai/sandboxes/forkread 2026-11-05 |
|
| Deployment & Data Residency | Full |
|
Sandboxes run in one chosen region (Oregon, North Virginia, London or Frankfurt), while agents, MCP servers and jobs spread across the regions a policy allows, and location policies stop any resource from running in a prohibited region. Private networking connects the customer's VPC through private endpoints, and a bring your own metal option runs the sandbox runtime on servers in the customer's own VPC, connected to Blaxel's control plane. Sourcedocs.blaxel.ai/infrastructure/regionsread 2026-11-05 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
Starting points come as hubs and templates. The MCP Hub offers prebuilt MCP servers, the Sandbox Hub ready development environments, and deployment templates and full repository examples start new agents and servers, with tutorials for Claude Code, Codex, OpenClaw and a dozen frameworks. These are tools and building blocks a developer assembles, not finished agents a business adopts. Sourcedocs.blaxel.ai/examples/overviewread 2026-11-05 |
|
| Triggers & Channel Coverage | Full |
|
Work starts on schedules, API calls and inbound requests. Sandbox schedules run a command on a cron expression, at a future datetime or after a delay, with an execution history for each run, and batch jobs run on cron schedules or on demand. Hosted agents and MCP servers answer at global endpoints, and a sandbox on standby resumes in about 25ms when a connection arrives. Sourcedocs.blaxel.ai/sandboxes/schedulesread 2026-11-05 |
|
| Model Flexibility & Routing | Full |
|
The customer chooses the models its agents call. The model gateway puts OpenAI, Anthropic, Mistral, Cohere, xAI, DeepSeek, Azure AI Foundry and Hugging Face models behind one endpoint with central credentials, policies can cap tokens, and any agent or coding tool running in a sandbox can use whichever model it is configured for. Sourcedocs.blaxel.ai/models/overviewread 2026-11-05 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
Every resource is managed through a REST API with public OpenAPI specifications, TypeScript, Python and Go SDKs and the bl CLI, covering sandboxes, snapshots, schedules, previews, agents, MCP servers, jobs, drives, volumes, policies, secrets and workspaces. An MCP server lets assistants such as Claude, Codex or Cursor manage Blaxel resources, and each sandbox exposes its own API and MCP server for files and processes. Sourcedocs.blaxel.ai/api-reference/introductionread 2026-11-05 |
|
| Testing, Debugging & Optimization | Not documented |
|
There are no test cases, scoring or quality gates for an agent's behavior. Traces and logs show what a run did after the fact, and reinforcement learning or benchmark runs that use the sandboxes score themselves in the customer's own code. Sourcedocs.blaxel.ai/llms.txtread 2026-11-05 |
|
| Browser / Computer-use | Not documented |
|
No browser or desktop for an agent to drive comes with the platform. Sandboxes run web apps behind preview URLs for people to view, and an agent that needs to browse installs and drives a browser itself inside the microVM. Sourcedocs.blaxel.ai/llms.txtread 2026-11-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Pay as you go with up to $200 in free credits, at $0.0000115 per GB of RAM per second of active sandbox compute.
Active compute per GB of RAM per second, plus snapshot, image and volume storage
Included quota
Every account starts at Tier 0 with 10 concurrent sandboxes. Topping up $20 a month moves an account to Tier 1 with 50 sandboxes and $50 to Tier 2 with 200, rising through nine tiers to more than 100,000. Cron jobs, model gateway requests, internal traffic and internet egress are included.
What is public
Blaxel publishes every compute, storage and network rate, the quota tiers and the support and HIPAA add on prices.
Billing mechanics
Sandboxes bill per second of active compute by allocated memory, with CPU scaling alongside, and stop billing compute when they suspend after a few seconds idle. Snapshots, images and volumes bill monthly per GB, and prepaid top ups both fund usage and raise the quota tier.
Cost watchouts
Standby is free of compute but each standby sandbox's memory and filesystem snapshot bills at $0.20 per GB a month, and volumes bill on provisioned size whether or not data is written. SAML comes with the $1,600 a month dedicated support add on, and a HIPAA BAA costs $250 a month.
Variable cost rationale
Cost follows the memory size and active seconds of every sandbox, job and MCP server, plus snapshot storage for every sandbox kept on standby, so busy fleets scale the bill directly.
Additional watchouts
A sandbox with 4 GB of memory costs about $0.17 an hour while active and nothing in compute while on standby, plus about $0.80 a month to keep its 4 GB snapshot ready.
Overage / add-ons
Usage draws down prepaid credits, and topping up credits raises the account's quota tier automatically.
Sales call required
Mixed (some tiers require a call)
Free / trial
Up to $200 in free credits, with a free tier of 10 concurrent sandboxes
Lowest paid plan
Pay as you go usage, with tier upgrades from a $20 monthly top up
Commercial notes
Baseten acquired Blaxel in September 2026, and existing API keys and credits remain valid. Teams running on it include Webflow, Shortwave, Framer, Stripe and Airbyte.
Key ambiguities
Custom deployments, including bring your own metal and private networking, are priced by quote.
Missing data
Custom deployments, private networking and bring your own metal have no public price.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…