Agents
Plex AI
The flagship coding agent. Prompt it with an idea and get back a runnable project with configs, routes, styles and README. Then keep directing it file by file. Ten iterations max per reply, verified against the tree before every summary.
One agent, four models
Each request tries the chain in order and advances only on retryable errors. Figures below are vendor-published; provider plans decide your real limits.
| Model | Provider | Capacity | Role |
|---|---|---|---|
| gemini-3.6-flash | Google AI | Free tier, daily reset | Primary |
| openai/gpt-oss-120b | Groq | 131K context · 65K out · ~500 tok/s | Fallback |
| openai/gpt-oss-20b | Groq | 131K context · 65K out · ~1000 tok/s | Fallback |
| google/gemini-3.8-flash | OpenRouter | Per-model billing | Fallback |
Eight tools, one tree
Everything the agent can touch, with collision checks and path-aware creation baked in.
listFiles
Full project tree with IDs, types and parents.
readFiles
Targeted file reads before any edit.
updateFile
Surgical rewrites of existing files.
createFiles
Batch creation with full relative paths; folders auto-created.
createFolder
Explicit empty folders and nested paths.
renameFile
Renames with collision checks against siblings.
deleteFiles
Recursive removal of files and folders.
scrapeUrls
Live documentation pulled into context.
Available everywhere you work
The same agent, surfaced wherever the work happens.
IDE
The full studio: explorer, tabs, ghost suggestions and the agent panel side by side.
Open the editor →
Chat
The sidebar companion for questions, explanations and guided edits with shared history.
Start chatting →
Cloud agents
Parallel fleets that build, test and review across repositories while you review.
Launch a fleet →
GitHub
Import any repo to start from real code, export any project when it is ready to ship.
Read the sync docs →