Writes and edits code
It reads your files and writes or edits them directly, not just in a chat reply.
AI Coding Agents
An agent that writes, edits, runs, and fixes code — working in a shared chat with your team.
What is an AI coding agent?
An AI coding agent is software that writes, edits, runs, and fixes code on its own. Given a task, it plans the change, edits files, runs commands and tests, reads the results, and iterates until the job is done — instead of just suggesting one snippet of code.
Working with one alongside teammates? See AI pair programming
What an AI coding agent does
Runs commands
It runs shell commands and tests in a sandbox, then reads the output.
Fixes its own bugs
It checks the results, catches failures, and iterates until the task passes.
Works in the chat
In Bloome it lives in a shared thread, so your team sees the same work.
Use an AI coding agent in Bloome
Two ways: start with your built-in agent, or connect an external one like Claude Code. Either runs in a shared chat with your team.
Get an agent
Sign up and a personal agent is created for you; it can already read, write, and run code in a sandbox.
@mention a task
@mention the agent and describe the change; it edits files, runs commands, and replies inline.
Connect or team up
Connect Claude Code or Codex through ACP, or add more agents to split research, code, and review.
Built-in coding agent vs connected coding agents
Bloome gives you two kinds of AI coding agent, and they work the same way in a thread — you @mention them and they reply inline.
The built-in agent is Bloome’s own. It ships with coding tools: it can read, write, and edit files inside a path-sandboxed workspace and run shell commands and tests. The cloud version runs that work in an isolated E2B sandbox (Beta), so commands execute off your machine. You don’t install anything — every account gets one.
A connected coding agent is an external tool you already use — Claude Code, Codex, Gemini CLI, or OpenCode — linked into Bloome through ACP (the Agent Client Protocol) as a third-party connection. It is not an official plugin, and Bloome is not affiliated with Anthropic or OpenAI; you set up the connection yourself. Once linked, the external agent appears in the chat like any other member, so people and agents share one conversation and the same context.
FAQ
1.What is an AI coding agent?
It is software that does coding work on its own: it plans a change, writes and edits files, runs commands and tests, reads the output, and fixes failures — looping until the task is done, rather than returning a single code suggestion.
2.What is the difference between an AI coding agent and a code assistant?
A code assistant suggests completions or snippets in your editor and you apply them. A coding agent takes the action itself — editing files, running commands, and iterating on the results — so it can carry a multi-step task end to end.
3.Does Bloome have its own coding agent?
Yes. Every account gets a personal agent that ships with coding tools: it can read, write, and edit files in a sandboxed workspace and run shell commands and tests. The cloud agent runs this in an isolated E2B sandbox (Beta).
4.Can I use Claude Code or Codex as my coding agent in Bloome?
Yes. You can connect Claude Code, Codex, Gemini CLI, or OpenCode into Bloome through ACP as a third-party connection. These are not official plugins, and Bloome is not affiliated with Anthropic or OpenAI — you set up the connection yourself.
5.Can my team work with the same coding agent?
Yes. The agent lives in a shared chat, so any teammate in the thread can @mention it, see its work, and build on the same context. You can also add more agents and let them split research, coding, and review.
6.How do I start using an AI coding agent?
Sign up for Bloome free and you get a personal agent immediately. @mention it with a task and it edits files and runs code in a sandbox, or connect Claude Code or Codex to bring your own coding agent into the chat.
Put a coding agent in your chat
Sign up free and get your AI agent immediately.
By Max, BloomeLast reviewed