Mirage

What is Mirage?

Mirage is a virtual terminal for AI agents from Strukto. Instead of defining a separate tool for every service, it mounts sources such as Amazon S3, Google Drive, Slack, Gmail and PostgreSQL under one virtual filesystem, so an agent reads them with familiar commands like ls, cat, grep and find. Around that sit virtual CLIs such as git, slack and ntn, a Bash-compatible shell with pipes and redirects, and runtimes that decide where each command runs. A scriptable policy engine can hide files per agent profile, refuse commands, ask for approval or cap output, and credentials can come from environment variables, dotenv, AWS Secrets Manager or 1Password. Mirage runs in your own app, server or browser tab through Python and TypeScript SDKs and a CLI. The docs report 52% fewer tokens per task than MCP tools across 200 tasks, based on the AgentVFS benchmark. It is self-hosted and open source under Apache-2.0, so there is no listed price; connected services bill their own API usage.

Use cases

• Give agents one filesystem over many services
• Search Slack, S3 and databases with grep
• Cut token use compared with per-service tools
• Restrict what each agent can see or run
• React to Slack or GitHub changes as file events
• Investigate Langfuse and Jaeger traces
• Run agent shells in a server or browser

Who it's for

• AI agent developers
• Platform engineers
• Teams building coding agents

Why use Mirage

• One shell interface replaces a separate tool definition for each service, which the docs report cut tokens per task by 52% versus MCP tools.
• Agents already know ls, grep and jq, so there is little new tool syntax to teach.
• Agent profiles and the policy engine let you limit what each agent can see and run.
• Components are composable, so you can use only the filesystem, or add CLIs and runtimes as needed.
• It is Apache-2.0 licensed and self-hosted, so data and credentials stay in your infrastructure.

Mirage features

• Unified virtual filesystem over S3, Drive, Slack, Gmail and databases
• Virtual CLIs such as git, slack and ntn
• Bash-compatible Mirage shell with pipes and redirects
• Scriptable policy engine for commands and file access
• Agent profiles controlling visibility and permissions
• Virtual runtimes from in-process to cloud sandboxes
• Credentials from env vars, dotenv, AWS Secrets Manager or 1Password
• File-change notifications for external events
• Python and TypeScript SDKs plus a CLI
• Runs in Node.js, a browser tab or as a FUSE mountpoint

Mirage integrations

• Amazon S3
• Google Drive
• Slack
• Gmail
• PostgreSQL
• Redis
• Cloudflare R2
• Google Cloud Storage
• Langfuse
• Jaeger
• Chroma
• 1Password
• AWS Secrets Manager

How to get started with Mirage

1. Install Mirage for Python or TypeScript, or install the CLI.
2. Create a workspace and mount sources, for example RAM, Slack or S3, each under its own path.
3. Set modes, profiles and policies to control what agents can read, write and run.
4. Give your agent a shell tool that calls the workspace, and let it use grep, cat and pipes across mounts.
5. Add runtimes or sandboxes where commands should execute.

Mirage limitations

• Self-hosted only, with no managed service
• Latency and usage charges follow the connected services' APIs
• The shell is Bash-compatible but differs from Bash in places
• Runtime file and network access differ by runtime

Mirage pricing

Mirage is open source, so you can use and self-host it for free.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

Mirage alternatives

The main Mirage alternatives listed on Debuly are E2B, Modal, AgentBox, SmolVM, smolagents and Arize AX. The table compares pricing and platforms; order is A–Z.

1. 

   AgentBox

   CLI and macOS menu bar app that moves your project into isolated VMs and runs Claude Code, Codex, OpenCode or Pi inside them in parallel.

   Developer platformsCodingAutomation and agents
   Visit AgentBox (official site)
2. 

   Arize AX

   Observability and evaluation platform for AI agents and LLM apps, with tracing, issue detection and experiments on a free plan.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Arize AX (official site)
3. 

   E2B

   Cloud sandboxes where AI agents run code safely, with Python and JavaScript SDKs and per-second usage billing.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit E2B (official site)
4. 

   Modal

   Serverless cloud platform for running AI inference, training, batch jobs and agent sandboxes, billed per second of compute.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Modal (official site)
5. 

   smolagents

   Hugging Face's minimal open-source Python library for building agents that write their actions as code.

   Developer platformsAutomation and agentsCoding
   Open sourceVisit smolagents (official site)
6. 

   SmolVM

   Open-source secure, persistent computer for AI agents, used as a sandbox for browser agents and computer-use tasks.

   Developer platformsAutomation and agents
   Open sourceVisit SmolVM (official site)

Mirage FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from strukto.ai · fetched 24 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.