Delegate the work. Keep control.
Flame's Agents Manager is a real control plane for AI work. Create, monitor, stop, schedule, and review tasks across every project, each one isolated, permissioned, and reviewable before it lands.
Live status, tokens, and tool calls for every running, scheduled, and merging task, with a concurrency limit you set.
Six things you can do with every task
Not a chat box: a board for running, watching, and shipping AI work at scale.
Create
Spin up a task from a prompt: pick the project(s), model, branch, and how it should land.
Monitor
Watch status, duration, tokens, and tool calls live as each agent works, streamed over WebSocket.
Stop & retry
Interrupt a run instantly, duplicate it, or requeue it; you're always in control.
Schedule
Run now, at a specific time, or after another job finishes. Chain dependent tasks into a pipeline.
Isolate
Each task can run in its own git worktree and branch, so parallel agents never collide.
Review & land
Inspect the diff, commit with AI, then merge, open a PR, or push, with auto-merge when clean.
From prompt to merged, safely
Every agent run follows the same reviewable path, so autonomy never means losing the plot.
Compose
Write a prompt, pick the project(s), model, branch, and how the work should land.
Isolate
Each task can run in its own git worktree and branch; parallel agents never collide.
Work
The agent edits files, runs commands, and tests, with every action gated by your permissions.
Review
Inspect the diff, ask questions in read-only mode, then commit with AI or by hand.
Land
Merge, open a pull request, or push, with auto-merge when the branch is clean.
It already knows what you're looking at
- Context attached automatically: files, diffs, terminal output, test results, even a live browser page.
- When the agent needs a decision, it asks: you answer with a tap, and it resumes.
- Tool cards show every edit, command, and test as it happens.
- Multi-project chat grants scoped read/write across several repos at once.
- Switch model and effort per conversation, mid-thread.
Claude, frontier APIs, or fully local
Use Claude's Fable, Opus, Sonnet, and Haiku, or connect OpenAI and Gemini, or run entirely offline with Ollama and LM Studio.
AI woven through the whole editor
The same intelligence powers completion, search, voice, and one-click actions, not just a side panel.
Voice control
Speak to navigate or hand a task to an agent. Everything else stays read-only.
Semantic search
Find code by meaning with a local embedding index, or use “find code like this” from any selection.
Inline completion
Fill-in-the-middle suggestions grounded in your project's symbols and nearby code.
Next-edit prediction
Cursor-style predictions of your most likely next change: jump and accept.
Predefined prompts
One-click actions: fix a failed command, review a diff, fix failing tests, resolve conflicts, build API requests.
Work summary
A standup-ready recap of commits, branches, projects, and tokens across your fleet.
Autonomy with a seatbelt
Every agent action is governed. You decide what's allowed, where, and whether to remember the choice.
- Global
- Group
- Project
- This chat
- File read / write
- Terminal & git
- Browser
- IDE control
- Allow once
- Deny
- Save a rule
- Ask every time
- IDE kill-switch
- Protected paths
- Command safety check
- Audit log (user vs agent)
Put an agent on it.
Join early access and let Flame's agents take the busywork while you keep the wheel.
No spam, ever. Just your invite and the occasional release note.