Continue when a program asks a question.
Answer setup prompts in the same terminal session, then inspect the resulting configuration. Try our small Python fixture before bringing your own installer.
Try interactive prompts →For teams automating Linux terminal work
A persistent terminal for your AI agent’s interactive work.
Give your agent a Linux environment that signals when an interactive program needs input. Keep terminal sessions between tool calls, answer prompts and continue work through MCP.
No credit card required. Plans from $39/month.
Bring your own agent and model access.
interactive setup
$ python3 interactive-setup.py
Project name: demo
↑ program waits for input
Create local configuration? [y/n]: y
↑ continue in the same session
Wrote demo-config.json
$ cat demo-config.json
{
"project": "demo",
"enabled": true
}Bring one workflow that needs manual recovery.
For automation consultancies and internal tooling teams already using AI agents: start with the recurring task that loses a prompt, session state or a result.
Recognize the problem
Where does a person have to step in?
- ? An installer waits at a prompt your agent cannot answer.
- ? A REPL or database session loses state between tool calls.
- ? A CLI test needs repeated manual navigation and result checks.
Evaluate the outcome
Keep your agent. Compare the workflow.
- ✓ Run the same approved task in an agend environment.
- ✓ Verify the output and count human interventions.
- ✓ Compare setup effort, completion and recovery with your current tools.
If an API or noninteractive script already completes the task reliably, keep using it. Evaluate agend where interactive state makes the work harder.
Start with the work you repeat.
Choose an approved Linux workflow where interactive state matters. Keep your current agent and compare the outcome.
Keep the state you need for the next step.
Reuse a Python variable or database session across tool calls. Verify that your agent returns to the existing process and handles its exit correctly.
Explore persistent sessions →Check the result as well as the interaction.
Evaluate recurring CLI tests with an owned fixture and explicit assertions. Record failures and manual interventions before deciding whether an agent improves the workflow.
Build a workflow comparison →A prompt, a process exit and a healthy application are different signals. Understand terminal readiness.
Your agent brings the reasoning. agend supplies the terminal.
Connect through the agend CLI to an isolated Linux environment with interactive terminal tools and persistent storage.
Input readiness
agend signals when an interactive program is waiting for terminal input. Your agent can read the prompt and respond. Task success still needs its own verification.
Isolated Linux environments
Each environment runs in a Firecracker microVM with its own guest kernel. Install the Linux dependencies your approved workflow needs.
Interactive tools through MCP
Read terminal output and send keystrokes from your existing agent. Keep using the same session for an installer, REPL or terminal application.
Persistent sessions
Keep session state between tool calls and files on persistent storage. Idle environments sleep; external connections may need to be re-established on wake.
Managed environments
Create an environment, install your dependencies and return to it for recurring work. Plan limits and available capacity determine what you can launch.
Bring your existing agent
Configure an MCP-compatible client with the agend CLI. Model access is separate. Validate your client, model and workflow together during the trial.
Connect your agent. Try one workflow.
1
Install
# One command. No dependencies.
curl -fsSL https://agend.sh/i | sh2
Authenticate & Configure
# After creating your account
agend login
agend env create
agend configThe CLI detects supported installed clients. Configure your chosen client, then test it with your workflow. Read the CLI setup guide, including Windows installation.
3
Operate
Ask your agent to:
1. List your agend environments.
2. Run the interactive setup example.
3. Answer the two prompts.
4. Read and verify the output file.Start with the reproducible installer example, then compare a task from your own work.
Plans for persistent terminal workflows.
Pay for the environment, not the AI. You bring your own agent — agend charges only for compute and storage.
Developer
$39/mo
- 1 environment
- Full root access
- 1 shared vCPU · 3 GiB RAM
- 10 GB storage per environment
- Interactive apps (vim, htop, psql)
- Sleeps after 30 min idle
- Persistent sessions and input readiness
3 environments
Pro
$79/mo
- 3 environments
- Full root access
- 2 shared vCPUs · 8 GiB RAM
- 50 GB storage per environment
- Interactive apps (vim, htop, psql)
- Sleeps after 2 hr idle
- Priority support
Enterprise
Custom
- Discuss your environment count
- Review CPU, RAM and storage needs
- Discuss deployment requirements
- Agree on support and service terms
7-day trial: one environment, 512 MiB RAM and 5 GB storage, with CPU capped at 25% of one core. Paid resources above are per environment and use shared CPU capacity. No credit card required.
Idle sleep starts after the client disconnects: 30 minutes on Developer, 2 hours on Pro. External connections may need to reconnect after wake. Environment launches depend on available capacity.
Join during early access and lock in founding member pricing — 30% off annual plans, forever.
Questions before you try a workflow.
Who is agend for?
Teams already using an AI agent for recurring Linux terminal work that needs interactive input or persistent session state. Bring an approved test workflow and compare completion, manual recovery and setup effort with your existing tools.
Does agend include an AI model?
You supply your agent and model access. agend supplies the remote Linux environment and MCP terminal tools. Model subscriptions and API charges are separate from your agend plan.
How do I connect my agent?
Create an account, install the CLI, run agend login, create an environment with agend env create, then run agend config to configure a supported installed client. See the setup guide for Linux, macOS and Windows.
Which clients can I use?
The CLI supports configuration for clients including Claude Code, Claude Desktop, Cursor, VS Code, Gemini CLI and others. MCP support and model behavior vary by client and version. Test your chosen client with your actual workflow during the trial.
What does the 7-day trial include?
One environment with 512 MiB RAM and 5 GB storage, with CPU capped at 25% of one core. No credit card is required. Start with a small workflow such as our interactive installer example. See paid plans for larger resources.
What does input readiness tell my agent?
It tells the agent that an interactive program is waiting for terminal input. A process exit is a separate event. Neither proves that a service is healthy or a remote transaction succeeded: verify the exit status, output file or application-specific health check. Read about the different signals.
How is agend different from other terminal tools?
Other tools also support PTYs and persistent sessions. agend combines interactive input-readiness signals, terminal tools through MCP and a managed Linux environment. The useful comparison is your task: does it complete correctly with less recovery or setup work?
What happens when an environment sleeps?
The client keeps its connected environment awake. After disconnect, idle sleep starts after 30 minutes on Developer and 2 hours on Pro. Files persist on the environment's storage. Processes pause during sleep, and external connections may need to reconnect after wake. Recheck application health before continuing work.
How is the environment isolated?
Each environment runs in a Firecracker microVM with its own guest kernel and filesystem. Your agent can install Linux packages and access reachable network services. Access to external systems still depends on the credentials and permissions you give it.
When should I use a script or API instead?
If a noninteractive command, script or API already does the job reliably and cheaply, keep it. agend is worth evaluating when interactive prompts, session state or recovery work make your recurring task difficult.
Test the task you keep recovering by hand.
Bring one recurring terminal workflow. Compare successful completion and manual recovery with your current setup during a 7-day trial.
brew install agend-sh/tap/agend