Connect your project and your AI.
Connect GitHub and choose your AI model. Open a personal or organization repository, or create a new one; MantisCode prepares the remote workspace.
Cloud Agents
Your agent.
In the cloud.
You're in charge.
AI agents that work on your GitHub repositories in a cloud workspace. Direct from your phone, review the diff and ship when you decide.
Coming soon to the App Store
Your way of working.
Connected.GitHubOpenCodeAGENT RUNTIMEOpenSpecSPEC-DRIVEN DEVELOPMENTYour repositories.
Your AI providers.
Connected.GitHubOpenCodeAGENT RUNTIMEOpenSpecSPEC-DRIVEN DEVELOPMENTYour repositories.
Your AI providers.
Your Workflow
Your phone directs.
The cloud runs.
An agent with access to your project's files, commands and tests works in a cloud workspace. You set the goal and decide what ships.
Direct the work.
Write or dictate the goal, add files and screenshots and answer the agent's questions. It edits, installs dependencies and runs commands and tests in the cloud.
Test, review and ship.
Open Preview, review the diff, ask for adjustments and decide when to commit, push or open the pull request.
What to Build
Start with a goal.
Turn it into code.
Fix a bug.
Describe the failure, attach a screenshot and ask the agent to find the cause, apply the fix and verify the result.
Review the fixBuild a feature.
Define the outcome and constraints. Organize the plan and tasks with Changes, implement with the agent and validate the work.
Discover the toolsTurn a design into code.
Explore an interface with prototypes, bring the design to the agent and ask it to implement it. Then check the web app in Preview.
From prototype to projectYour Environment
Everything to take your idea
all the way to the pull request.
Context, planning, design and review alongside the agent. Tools connected to the same project to move from the first goal to delivery.
01 — CONVERSATION + CONTEXTSet the direction.
Set the direction.
Follow every step.
Talk or write, mention files with @ and attach a screenshot. Watch the tools in action, answer questions and grant permissions when the agent needs them.
The agent asks. You decide.
02 — CHANGES + PROTOTYPESThink first.
Think first.
Build with intent.
Organize specifications, tasks and validation with Changes. Explore interfaces on a prototype canvas and bring the design into the conversation with # to ask for its implementation.
FROM PLAN TO TASKS
03 — PREVIEW + DEVICESTest your site.
Test your site.
On any screen.
Browse and edit files. Open your web application's preview with the development server running and check how it adapts to mobile, tablet and desktop.
04 — REVIEW + DELIVERYEvery line.
Every line.
Under your eyes.
Review modified files, mark what you've already seen and send comments about specific lines to the agent. Ask for the changes you need and decide when to commit and push, open a pull request or merge.
See the full workflowDiffCommitPushPRMerge
You Decide
The agent moves.
You stay in charge.
You don't need to watch the screen while it works. It notifies you when it needs you, you can correct course on the fly and go back if the result doesn't convince you.
It lets you know when it needs you.
Get a notification when the agent finishes, asks you a question or needs a permission. Tap it to return to the session. The notification doesn't include the response content.
Push notificationsCorrect it without stopping it.
Write while the agent works to redirect the task. If the turn ends before applying your instruction, it's marked to be resent.
Send while it worksGo back to an earlier point.
Choose a reply and use Undo from here. Before confirming you see which files will be reverted; that interaction and the following ones leave the conversation.
Undo from hereShow it before publishing it.
Create a temporary link to Preview so someone else can try your app. Anyone with the link can use it and it can't be revoked: it expires when the preview expires or the workspace hibernates.
Share previewExtensions
Your agent.
Your tools.
Choose your models
Choose the model and reasoning for each way of working. You can also connect your AI providers through OAuth or an API key, depending on the available options.
Extend with Skills
Import skills from a GitHub repository and use them in the conversation with /. The ones your repository already includes come with the project.
SSH Servers
Your server.
Your agent in action.
Take the agent to the servers you already use. Connect over SSH and ask it to investigate a failure, check a service or run a task from the conversation on your phone.
Understand what's happening
Check logs, processes and services. Ask for a diagnosis and an explanation before deciding the next steps.
Delegate specific tasks
Request maintenance or a deployment following your project's procedure. State the destination, the goal and what you want the agent to do.
Define the scope
The agent acts on the server you choose, with the permissions of the remote account you've connected.
Example request
“On the server My site, check why the application returns a 502 error. Explain the diagnosis and the logs that support it before changing anything.”
DiagnosisLogs and servicesMaintenance
Save a connection with a recognizable name and use it in your requests to say where the agent should act.
Add the server in Settings → SSH Access. Install the public key and verify the server identity before using it.
Your Plan
Start with a site.
Build your next app.
Choose the workspace capacity your project needs. Sites and Apps share the tools; you choose how many resources to work with.
Monthly or yearly. Check prices and availability for your account and device in the app.
FOR YOUR NEXT SITE
Sites
Simple websites and frontend projects.
Landing pages, portfolios and sites to shape your idea.
Base workspace capacity
- 1 vCPU
- 2 GiB RAM
- 10 GiB disk
MORE RESOURCES TO BUILD
Apps
Web or hybrid applications.
Projects with a frontend, a backend and a more demanding development environment.
Base workspace capacity
- 2 vCPU
- 4 GiB RAM
- 20 GiB disk
On both plans
- Cloud agents and GitHub repositories
- Changes, Skills and prototypes
- Files, tests and web Preview
- Review, commit, push and pull request
- SSH access to your servers
Access to your AI provider is configured separately from the MantisCode subscription.
Questions
The essentials,
answered.
What do I need to get started?
An active MantisCode subscription, your GitHub account and an AI model available in the app. Check the plans there and choose an existing repository or create a new one.
Where does my code run?
In a cloud workspace: a remote, isolated environment where the agent works with the files and runs the project's commands and tests. It has limited resources, hibernates after a period of inactivity and needs time to recover when you return.
What happens to my work if I stop using the app?
The workspace hibernates after 60 minutes without activity and recovers when you return. If it stays hibernated for 7 days, it is destroyed with all its content; also when you sign out or delete your account. What you've uploaded to GitHub with commit and push is kept; what isn't may be lost.
Do I need to leave my computer on?
No. Work runs in the MantisCode workspace, without depending on your computer. From your phone you give instructions, follow progress and review the result. You need an internet connection to work with the remote environment.
Which models can I use?
Choose from the models available in your workspace or connect your providers with OAuth or an API key, and set a model per work mode. The MantisCode subscription is managed separately.
How does the subscription work?
Sites and Apps plans have monthly and yearly options. Check prices, availability and applicable terms in the app. Purchases and restores are managed from the app and the store associated with your device.
Start in the cloud.
Follow along from your phone.
Your repository, your models and a development environment with agents. Check the plans in the app, connect GitHub and choose a model to get started.