Cendra
Cendra is your AI teammate in Cendriix — a conversational assistant that doesn’t just answer questions, it does the work. Ask it to explain a codebase, fix a bug, build a feature, run tests, or open a pull request, and it carries the task end-to-end while you watch the work happen live. Access via Operate → Cendra in the sidebar.
What you can do
- Chat in plain language — Ask anything about your repositories, agents, or workspace.
- Build and fix — Describe an outcome (“add a logout button”, “fix the failing checkout test”) and Cendra plans the work, writes the code, runs it, and proposes a PR.
- Use your voice — Dictate a message instead of typing.
- Pick up where you left off — Every conversation is saved as a thread you can revisit, search, and pin.
How it works
You type (or speak) a message in the composer at the bottom. Cendra streams its response back in real time, and as it works, the right-hand panel updates to show exactly what’s happening — the agents it’s running, the files it’s writing, the commands it executes, and the pull request it’s preparing. You stay in control: review the changes, ask follow-ups, and approve the PR when you’re happy.
The composer
The message box is where every request starts. It supports:
- Modes — choose how Cendra approaches your request:
- Auto — Cendra picks the right depth for you (default).
- Fast — quick answers for simple questions.
- Deep — more thorough reasoning for complex problems.
- Build — go straight to building and shipping code.
- Voice input — tap the mic to dictate.
- Attachments — add files for context.
- Slash commands — start a message with a shortcut:
/build,/fix,/test,/explain,/status. - @ mentions — point Cendra at something specific:
@agent:,@repo:,@workflow:,@mcp:,@kb:.
Conversation history
The left panel lists your conversations grouped by date, with search and pinning. Each thread shows its status at a glance — Executing, Done, or Failed — so you can track long-running work even after you close it.
The activity panel
While Cendra works, the right panel gives you a live window into the job. Switch tabs to follow different parts of the work:
| Tab | What it shows |
|---|---|
| Agents | The team of agents working on your task and what each one is doing right now. |
| Log | A step-by-step trace of reasoning, tool calls, and command output for each turn. |
| PR | The pull request being prepared — title, branches, CI checks, a self-review, the diff, and a merge button. |
| Browser | Screenshots of the app as Cendra navigates it during testing. |
| Files | The code artifacts being written, with diff previews you can mark as reviewed. |
| Terminal | Command output from the sandbox where your code runs. |
| Studio | An inline editor preview, with a one-click Open in Studio for the active repo. |
| Grid | Device-testing progress and QA screenshots when Cendra runs UI tests. |
Active runs
Long-running builds keep going even if you navigate away. A live indicator tracks progress, and active runs appear in the sidebar so you can jump back into any in-flight job.
Grounded, not guessing
Cendra answers using Cortex — your workspace’s knowledge layer — so responses about your code come with file-level citations instead of guesses. When it builds, it uses your connected Agents, Flows, Skills, and Connectors.
See also: The Studio, Cortex, Agents, Cockpit