Amika

Amika is infrastructure for running coding agents and human developers inside the same cloud workstations. It gives each job a sandboxed virtual machine with the repository, services, tools, seeded data, and coding agent it needs already in place. Those workstations can serve as everyday cloud devboxes, interactive spaces where a team steers background agents, or execution environments for automated software workflows. The practical goal is to replace fragile local worktrees and one-off agent setups with repeatable environments that a person or an agent can open from anywhere.

Amika calls these environments Rigs or sandboxes, depending on the part of the product. Their configuration lives alongside the code in a repository, so the base image, services, startup behavior, and agent choices can be reviewed and versioned like other project files. A Rig can clone one or more repositories and run supporting services such as a database or web server. It can also expose ports through HTTPS preview URLs. Snapshots preserve the installed tools, running services, and warm caches, which lets a stopped environment resume without rebuilding the whole development stack each time.

Developers can connect over SSH, place a sandbox on a VPN, or work through Amika's web interface. The same environment can be controlled from the CLI, HTTP API, and supported work tools. The site shows entry points through Slack, Linear, and GitHub, including assigning a ticket, responding to a failed check, or asking an agent to address review comments. Codex, Claude Code, and OpenCode are named as supported agent harnesses, and customers bring their own model credentials. Amika does not add a markup to the model tokens those providers bill.

Collaboration is a central part of the product. Teammates can enter the same live sandbox, inspect an application through a preview URL, watch an agent session, redirect it, or take over the work. That is useful when a background agent reaches a judgment call or when a change needs hands-on verification before it moves forward. Sessions can be retained, while higher plans add more storage and access to past sessions. The environment can also run on Amika-managed infrastructure, a compatible cloud or Kubernetes setup, or hardware a customer controls, although bring-your-own infrastructure is positioned as an Enterprise capability in the current pricing.

The strongest fit is an engineering team that already uses coding agents and needs them to operate against a realistic stack. A background agent can pick up a ticket, reproduce a production issue with logs, run tests and type checks, then prepare a pull request if the checks pass. A team can also standardize a development environment once and give it to both people and agents. Amika is different from an agent editor or model provider because it concentrates on the machine, network access, reproducibility, orchestration, and collaboration around the agent. It can use several agent harnesses instead of replacing them with a proprietary coding interface.

For deeper automation, Amika lets teams define event-driven workflows in code and start them from webhooks, schedules, or API calls. The examples include responding to Sentry alerts, collecting matching logs from GCP or AWS, reproducing a failure, running checks, and opening a pull request only when the result is ready. Teams can make tests, type checks, or custom validators part of that path and keep a person available to review or redirect the run. The product also records agent sessions with their transcripts, tool activity, and token costs, which gives a team material to inspect when it is refining prompts, rules, budgets, or rate limits. This is more operational work than opening a coding assistant in an editor, but it supports repeatable jobs that continue in the background.

Amika has a free plan for one seat with two concurrent sandboxes, 25 stored agent sessions, and a small amount of sandbox credit. Paid Developer and Team plans raise concurrency and session limits, add multiplayer and organization features, and include more compute credit. Sandbox compute, memory, and storage are metered, so a subscription does not make infrastructure usage unlimited. Enterprise customers can request custom infrastructure, security controls, support, and deployment options. The open-source CLI can also run local Docker-based coding agent containers, but the hosted control plane and its collaboration features are the broader commercial product.