Installation
FreeCloud platform for coding agents
FreeFree tier
About Installation
Ellipsis is a cloud platform designed to run coding agents safely and at scale. It provides each agent with an isolated ephemeral sandbox, scoped GitHub permissions, hard budget caps per session/day/month, and full observability across the fleet. Agents can be triggered by events (15 types across GitHub, Linear, Slack, and Sentry), scheduled via cron, or invoked on demand with @ellipsis mentions. Agent configurations are defined in YAML and deployed with git push, supporting version control and code review. Ellipsis integrates with AI agents like Claude Codex and Gemini, and has been adopted by over 400 engineering teams on 67,000+ repositories. Pricing is usage-based with no per-seat fees, and new organizations receive a $100 credit.
Key Features
Each agent session runs in an isolated, ephemeral sandbox with repos cloned and toolchain preinstalled
Scoped GitHub permissions narrowed at token mint and repositories scoped by name
Hard budget caps enforced per session, per day, and per month before a sandbox exists
Full observability with a unified pane for monitoring active agents, shipped work, and cost
Event-driven agents: 15 event types across GitHub, Linear, Slack, and Sentry trigger sessions automatically
Scheduled agents via cron expressions defined in YAML
On-demand agents via @ellipsis mentions on GitHub, Slack, or Linear
Agents defined as code in YAML: prompt, triggers, and spend caps version-controlled with git push
Warm cached images for dependency installs to start repeat sessions in seconds
Sized resources: CPU, memory, and timeout set per agent in YAML from light triage to heavy builds
Pros & Cons
Pros
- No per-seat fees; pay only for actual usage (tokens, CPU, memory)
- Isolated sandboxes ensure parallel work on same repo without conflicts
- Fine-grained permissions and budget caps provide security and cost control
- Full observability and session replays help audit agent actions
- Supports multiple agent providers (Claude Codex, Gemini) and custom agents
- Event-driven and scheduled agents enable unattended automation
- Agents as YAML config integrates with existing Git workflows
- New organizations get a $100 credit to start
Cons
- Pricing is usage-based and can vary; heavy usage may be costly
- Cloud-dependent; no self-hosted option mentioned (besides BYOK/Bedrock but compute still on Ellipsis)
- Limited to specific agent providers (Claude Codex, Gemini) unless custom integration
- Requires setup of YAML configuration and integration with GitHub, Slack, etc.
- No free tier beyond the initial $100 credit
Best For
Automated code review on pull requests with inline commentsFix production alerts from Sentry (e.g., add null checks, open PR before team wakes up)Bump dependencies and fix breakage across call sitesAdd missing tests for coverage gapsTrack down and fix flaky testsBuild small features end-to-end (API routes, schema, validation, tests)Generate daily standup reports for engineering teamsManage stale PRs and automate routine tasks
FAQ
How does the $100 credit work?
Every new organization starts with a one-time $100 credit. It applies to usage automatically and covers roughly 135 code reviews or 50 small features before you pay anything.
What does usage-based mean here?
Every agent session runs in its own sandbox, and you pay for exactly what that session consumes: CPU and memory metered by the second, tokens passed through at Claude's rates with no markup, and a per-session platform fee equal to 100% of the session's token usage. Nothing runs idle, so nothing bills idle.
So what does a typical task actually cost?
A quick fix like a 2am Sentry alert costs about $0.69 total. A code review on a medium PR is about $0.74. A flaky test investigation is about $1.48. Bigger tasks with more files and longer sessions cost more.
Do I need to pay per developer?
No. There are no per-seat fees. Your entire team can use Ellipsis. You only pay for actual usage.