Ellipsis

Ellipsis is a platform for deploying managed AI coding agents that review pull requests, build features, and investigate production issues. Teams define agents in YAML in their repo and run them in isolated cloud sandboxes with permissions, budgets, and tracing.

AI Agent Development AI Developer Tools Programming Agent

Visit Website

Overview

Ellipsis is a platform for deploying AI software engineers as managed Claude Code agents. Teams define agents in YAML files in their repositories, then let Ellipsis run them to review pull requests, build features, investigate production issues, and carry out other coding tasks.

The product combines cloud execution, permissions, budgets, observability, and tracing in one control plane. Agents run in isolated sandboxes, can be limited to specific repos and branches, and leave an auditable record of what they did and how much they cost.

Features

Agent config as code

Define agents in YAML with fields for prompt, model, triggers, tools, and spend caps, then store that config in the repo and version it like code.

Multiple run triggers

Start runs from schedules, GitHub or Linear events, @mentions, the CLI, or the REST API so agents can fit into existing developer workflows.

Managed cloud sandboxes

Run each agent in an isolated sandbox, with the cloud control plane handling permissions, status, and run bookkeeping.

Permissions and budget controls

Set per-agent allow-lists for tools and MCP servers, scope agents to specific repositories and branches, and apply run, daily, and monthly spending caps.

Observability and tracing

See run history, statuses, costs, outputs, thinking, tools used, tests run, and shipped diffs for auditing and debugging.

Usage-based billing

Use pricing based on tokens, CPU, memory, and a platform fee, with a free monthly usage allowance and no per-seat or idle charges.

Common use cases

Automated PR review
Create agents that review pull requests against team conventions and leave inline comments before code merges.
Issue triage and debugging
Build agents that inspect alerts or failing tests, trace likely causes, and open a fix plan or pull request.
Recurring team operations
Use scheduled agents to post standup digests, summarize activity, or keep documentation and routine workflows up to date.
Small feature implementation
Deploy feature-building agents that edit code in a sandbox, run tests, and open a PR for the result.
Governed team rollout
Manage org-wide agent usage with repo and branch scoping, spend limits, and visibility into active runs and costs.

Pros and Cons

Pros

Agents are defined in repo-local YAML, which keeps behavior versioned alongside code.
Runs can be triggered from GitHub, Linear, schedules, @mentions, the CLI, and the API.
Cloud sandboxes, least-privilege controls, and repository or branch scoping help constrain agent access.
Tracing and run history make agent behavior and cost visible to the team.
Pricing is usage-based, with a free monthly allowance and no per-seat or idle fees.

Cons

The documentation provided here is partial, so some setup and integration details are not fully visible in the source.
Private Cloud setup is described as taking about an hour, so it is not fully instant like the hosted cloud option.

FAQ

How do you set up an Ellipsis agent?

Ellipsis is set up from a YAML config in your repository. The docs describe defining agents in YAML, then deploying them from GitHub, the CLI, or by mentioning @ellipsis.

What can trigger an Ellipsis run?

The source shows runs can start on a schedule, on GitHub or Linear events, or by mentioning @ellipsis. The docs also reference a REST API and CLI for starting runs.

What kinds of tasks does Ellipsis handle?

Ellipsis can post PR review comments, build or update code in a sandbox, investigate issues such as flaky tests or production alerts, and open pull requests with the result.

How does Ellipsis pricing work?

The pricing page says every account gets $10 of free usage each month, and public GitHub repositories are free. It also says there are no per-seat fees or idle charges.

How does Ellipsis handle security and permissions?

The security and cloud pages say agents run in isolated sandboxes, with least-privilege permissions, configurable allow-lists, and zero source code retention on Ellipsis cloud.