Run popular coding agents in sandboxed cloud VMs. Deeply integrate with your existing tools for ultimate context. Turn every teammate into a power user.
Deprecate legacy users route
Remove deprecated /v1/legacy_users endpoint and migrate all services to use /v1/users for consistency.▊
Acceptance Criteria
- Remove /v1/legacy_users endpoint from all backend services
- Update all clients to use /v1/users
- Verify new endpoint is working
- No references to legacy endpoint remain in codebase▊
Remove deprecated /v1/legacy_users endpoint and migrate all services to use /v1/users for consistency.▊
Acceptance Criteria
- Remove /v1/legacy_users endpoint from all backend services
- Update all clients to use /v1/users
- Verify new endpoint is working
- No references to legacy endpoint remain in codebase▊
Agents your developers trust. Don't settle for less.
Claude Code
Codex
Gemini CLIFire off. Forget. Come back to results. Every task runs in an isolated remote sandbox. Agents can clone as many repos as needed.
Commands. Quick actions for common tasks. Share with the entire team. Reference tickets, repositories, and more.
Integrate your tools. Linear, GitHub, Slack, and more. The ultimate context. Perform actions across them.
Configure MCPs. Automated browser testing. Investigate error traces. Connect a database. No more local configuration burden.
Take over as needed. SSH from your local machine or use VSCode in the browser. Make surgical changes when agents are not enough.
A Linux sandbox in the cloud for every task. Let agent's work to their full potential.
Blazing fast Micro-VMs
Sandboxes spin up in under 5 seconds. Get responses fast.
Secure by default
Ephemeral environments that spin up when you need them, down you're done.
Persistent filesystems
Restart a machine with the exact state where you left it.
Connect via SSH
Connect to your agent's sandbox from the terminal, or with an IDE SSH plugin.
Complete isolation
Work on many tasks concurrently without conflicts.
Take over locally
Move an agent's environment to your local machine when needed.
Centralize agent configurations. Scale to the entire team. Enable non-technical members. Turn every teammate into a power user.
Reusable prompts
Define once for your team. Use again and again.
MCPs
Connect your database. Retrieve error telemetry. And more.
Connect subscriptions
Claude Code supported. Codex coming soon.
Managed configuration
Offload configuration burden. Enable your team to focus on work.
Invoke from anywhere. Return to results. Blocks can do almost anything across connected tools, just ask.
Thread to PR. Unblock teammates with technical answers from your codebase, create tickets from threads, and more.
A teammate inside your repos. Ask Blocks to review PRs, write tests, manage issues, update pipelines, and more.
The most technical PM ever. Iterate on tickets with Blocks until perfection, then ask Blocks to implement it.
Create custom agents with the Blocks SDK. Any framework, any agent. Respond on Slack, Linear, GitHub, or the dashboard.
from blocks import agent, on
from smolagents import CodeAgent, WebSearchTool, LiteLLMModel
@agent(name="custom-slack")
@on("slack.mention")
def custom_slack(input, config: SmolAgentConfig):
model = LiteLLMModel(
model_id=config.model,
)
agent = CodeAgent(tools=[WebSearchTool()], model=model)
client = slack_sdk.WebClient(token=SLACK_TOKEN)
client.chat_postMessage(channel=channel, text="Agent is thinking...", thread_ts=ts)
for _, step in enumerate(agent.memory.steps):
messages = step.to_messages()
...We'll work with you through the entire setup process. A Blocks engineer will provide education, configuration, and support.
Direct line of communication
Have a question? Call or send a message in our community Slack.
Hands-on training
We'll provide small group training and support.
FAQs
Blocks focuses purely on background agents, and is not used synchronously in an IDE like Cursor, Claude Code, and so on. Blocks orchestrates background coding agents in containerized sandbox environments where they can interact with connected tooling. Think Devin, but instead of a proprietary agent, you can use familiar agents like Codex or Claude Code in a Devin-like manner.
We're currently in beta and free for early users (with limited quotas). Pricing will be announced before general availability, with plans designed for both individual developers and teams.
Yes, you can use your existing coding agent subscriptions with Blocks. We currently support Anthropic subscriptions for Claude Code, and will be adding support for OpenAI Codex soon. Otherwise, you can use API keys for any agent if preferred.
Blocks connects seamlessly with popular development tools like GitHub, Linear, Slack, and more through secure OAuth integrations. Agents can read context and take actions across connected tools.
Like all coding agents, Blocks agents will only perform what they are requested to, but may fulfill requests in unexpected ways if they are not appropriately bounded (ex: agent may commit against a branch when the requester expected a PR since it was not explicitly mentioned). It is important to specify exactly what the agents behavior should be to the best of your ability to avoid unexpected behavior.