Ellipsis Review 2026: Cloud Coding Agents, Pricing & Limits

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 13, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Ellipsis is a cloud platform for running coding agents. You define an agent in a YAML file in your repository, deploy it by merging, and Ellipsis runs it in an isolated cloud sandbox when something happens: a pull request is opened, a Sentry alert fires, someone mentions it in Slack or Linear, or a schedule comes due. The agent does its work with Claude Code or Codex and reports back, usually as a pull request or a comment.

The company started in 2023, went through Y Combinator’s Winter 2024 batch and first became known for automated pull request review. It now describes itself as an “Agents as a Service Platform” for feature work, background tasks and code review. Ellipsis is aimed at engineering teams that want to run Claude Code or Codex automatically and safely without maintaining their own runners, secrets and logging. This is an editorial research review based on Ellipsis’s website, pricing page and documentation. We did not run our own benchmarks or timed tests.

Quick verdict

Ellipsis is a focused infrastructure layer for teams that already trust Claude Code or Codex and want them running in the background with guardrails. Its strongest points are budgets, scoped credentials, audit logs and configurable multi-reviewer code review, all defined as code. In return, it supports only two agent harnesses, it is built around GitHub, its billing is entirely usage-based, and its published enterprise compliance story is thinner than some rivals’.

Pro: Agents, environments and code reviewers are defined in YAML and deployed through normal pull request review.
Pro: Detailed spend controls, with caps per session, per agent, per developer and per account.
Pro: Scoped short-lived credentials, no source retention after a session, and a full audit trail.
Pro: No seat fees, and sessions you start yourself can run on your own Claude or ChatGPT subscription.
Con: Only Claude Code and Codex harnesses are supported today.
Con: The documentation covers GitHub only. We found no GitLab, Bitbucket or Azure DevOps support, and Jira is listed as not yet available.
Con: Usage-based billing for tokens and compute is harder to budget than a flat seat price.
Con: SOC 2 Type I is certified and Type II is in progress. We found no published SSO or SAML details.

Visit Ellipsis

What is Ellipsis?

Ellipsis gives coding agents a managed place to run. Each session gets its own cloud machine with your repositories cloned in, your dependencies and tools installed, and credentials limited to the repositories you allow. Sessions can start from the dashboard, the el CLI, a REST API, Python and TypeScript SDKs, GitHub, Slack, Linear or Sentry events, or a schedule. You can follow a session live, send follow-ups and download the raw agent output. Ellipsis says it is installed on more than 67,000 repositories across more than 400 companies. That is the company’s own figure, and we have not verified it.

Key features

Agents as code

An agent is a YAML file that defines its trigger, prompt, environment, harness, model and budget. Changes deploy when you merge, so agent behavior goes through the same review as application code. You can also create and update agents through the API. Supported triggers include cron schedules, GitHub pull request, push, issue, CI check, workflow run and release events, new Linear issues, Slack events and Sentry alerts.

Cached environments

Environment files define repositories, variables, compute size and three setup hooks: dependency installs, builds after checkout, and a step before each start. Ellipsis caches each hook’s result and reruns only what a change affects, so sessions do not reinstall everything every time.

Configurable code review

A code_review file can define up to eight reviewers. Each reviewer has its own prompt, model, harness and path filters. Ellipsis runs the reviewers on each push and posts one combined GitHub review. Reviewers mark comments as blocking or not, and you choose whether a passing review stays silent, comments or approves, and whether a failing one comments or requests changes. An organization-wide file can cover every repository, with per-repository overrides.

Claude Code and Codex harnesses

Ellipsis runs the actual Claude Code and Codex agents. Current models include Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, plus GPT-6.1 Sol and other GPT-6 and GPT-5.6 models. Switching harness or model is a configuration change. Other agents such as Copilot, Gemini, Cursor and OpenCode are listed as not yet available.

Budgets and governance

Each session has a spend cap, which defaults to $250. Agents and accounts have rolling daily, weekly and 28-day limits, optionally split per developer. Credentials are minted per session and revoked with the sandbox, code is deleted with the sandbox, and you can set session data retention anywhere from zero hours to one year.

Who should use Ellipsis

Ellipsis suits GitHub-based teams that already use Claude Code or Codex and want them working in the background, reviewing pushes, investigating alerts or running scheduled maintenance. It also suits platform teams that would rather configure budgets, secrets and audit logs than build them around their own CI runners.

Who should skip it

Skip Ellipsis if your code lives outside GitHub, if you need agents other than Claude Code or Codex, or if procurement requires SOC 2 Type II and SSO today. Individuals who mainly want help in their editor will get more from an IDE assistant.

Strengths and limitations

Strengths

Governance built in. Budgets, scoped credentials, audit trails and configurable retention cover the controls teams usually have to build around self-run agents.
Reviewable configuration. Because agents, environments and reviewers are YAML in Git, changes are versioned and reviewed like code.
Flexible code review. Several specialized reviewers, each with its own model and path filter, can feed one GitHub review with clear pass and fail rules.
Transparent cost model. No seat fees, published per-token and compute rates, and the option to use existing subscriptions for hands-on sessions.
Runs in your cloud. The AWS deployment option keeps sandboxes and the control plane in your own VPC.

Limitations

Narrow platform support. The docs describe only the GitHub App for source control. Teams on GitLab, Bitbucket or Azure DevOps are not covered, and Jira is not yet available.
Two harnesses. You are limited to Claude Code and Codex. Some features also depend on billing: on the Claude Code harness, skills are not supported, and images and custom settings require credits rather than a Claude subscription.
Variable costs. Background agents always use credits, and compute is charged while a sandbox waits for input. Budgets help, but monthly spend still depends on how busy your agents are.
Compliance still developing. SOC 2 Type II is in progress, and we found no published SSO, SAML or SCIM details. Larger organizations should ask before committing.
Setup effort. Getting value takes YAML authoring, environment tuning and prompt work, and there are no in-editor completions or chat.

Pricing (as of October 2026)

Plan	Price	What you get
Managed SaaS	No seat fees. Usage-based, from prepaid credits.	Model tokens at Ellipsis’s listed per-model rates, a 10% platform fee, and organization compute at about $0.142 per CPU core-hour and $0.024 per GiB-hour. Credits do not expire. All features are included.
Personal use	Usage-based	Compute for personal accounts is free. Sessions you start yourself can be covered by a connected Claude or ChatGPT (Codex) subscription.
Your AWS account	10% platform fee, plus your own AWS costs	The control plane and agent sandboxes run in your AWS VPC. Tokens and compute are billed by AWS. All features are included.
Support packages (optional)	$5,000, $10,000 or $15,000/month	First response within 48, 24 or 4 hours, with a named support engineer on the top tier. Ellipsis says no support package is needed to use the platform.

Scheduled and event-triggered agents and API-key sessions always use credits, not subscriptions. Pricing and model rates can change, so confirm current details on Ellipsis’s pricing page and pricing docs. See Ellipsis pricing

Prices verified on October 13, 2026 against the official Ellipsis pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Ellipsis runs Claude Code and Codex agents in cloud sandboxes when events happen, so it fits best for recurring, well-defined work on GitHub repositories. These scenarios are based on Ellipsis’s documentation and the research behind our review, not on hands-on testing.

Layered pull request review

A team wants separate reviewers for security, database migrations and frontend conventions instead of one generic bot. One Ellipsis code_review file can define up to eight reviewers, each with its own prompt, model and path filters, and posts one combined GitHub review with blocking rules. Each reviewer runs as a metered session, so a busy repository with many reviewers increases spend.

First response to production errors

An on-call engineer wants a head start when an alert fires. A Sentry alert can start an agent that investigates the failure in a sandbox with the repository and dependencies ready, then comments its diagnosis on the Sentry issue. In our assessment, this works best as triage, with a human still deciding what to fix.

Turning Slack or Linear requests into pull requests

A product team wants small fixes requested in Slack or Linear handled without anyone opening an editor. Mentioning the agent in a thread or on an issue starts a session that can open a linked pull request and post the result back. Code still goes through normal GitHub review, and Jira is not yet supported.

Scheduled maintenance with a spend cap

A platform team wants a weekly dependency or cleanup pass with a hard spending ceiling. A cron-triggered agent with daily, weekly and monthly limits fits this pattern, and alerts fire as limits approach. Scheduled agents always use credits rather than a personal Claude or ChatGPT subscription.

Alternatives

CodeRabbit: a dedicated AI pull request review service.
Factory: a platform for software development agents aimed at teams.
Claude Code: Anthropic’s coding agent, which you can also run yourself in CI.
Warp: a terminal and agent platform with cloud agents and Factories.

Final verdict

Ellipsis has grown from an AI code reviewer into a managed runtime for Claude Code and Codex agents, with the budgets, permissions and audit trails that teams need before letting agents run unattended. It makes the most sense for GitHub teams that are already committed to those two agents. Teams on other platforms, or with strict enterprise compliance requirements, should wait or look elsewhere.

Try Ellipsis

FAQ

Is Ellipsis free?

Not entirely. Ellipsis charges no seat fees and compute for personal accounts is free, and sessions you start yourself can run on a connected Claude or ChatGPT (Codex) subscription. Scheduled and event-triggered agents use prepaid credits for model tokens, organization compute and a 10% platform fee. Check Ellipsis’s pricing page for current rates.

Is Ellipsis still an AI code review bot?

Code review is still a core feature, but Ellipsis says it has expanded from automated review into a platform for running coding agents for features, background tasks and review. Older articles that describe it only as a review bot may be out of date.

Which AI agents and models does Ellipsis support?

Ellipsis runs two harnesses: Claude Code, with Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, and Codex, with GPT-6.1 Sol and other GPT-6 and GPT-5.6 models. Copilot, Gemini, Cursor, OpenCode and others are listed as not yet available. Our Claude Code and OpenAI Codex reviews cover those agents.

Does Ellipsis work with GitLab or Bitbucket?

We found no support for them. Ellipsis’s documentation describes a GitHub App for source control, plus Slack, Linear and Sentry integrations, and lists Jira as not yet available. Teams on other platforms should ask Ellipsis directly.

Does Ellipsis keep my source code?

Ellipsis says your code lives only in the session sandbox and is deleted with it. Credentials are minted per session and scoped to the repositories you allow. You can set session data retention from zero hours to one year, and an audit trail of commands and file changes is kept. Ellipsis lists SOC 2 Type I as certified and Type II as in progress.

Ellipsis vs CodeRabbit: which should I choose?

They overlap only on review. CodeRabbit is a dedicated pull request review service, while Ellipsis is a general platform for running Claude Code and Codex agents, with code review as one configurable use. In our assessment, choose CodeRabbit if review is all you need, and choose Ellipsis if you also want event-driven and scheduled agents under the same budgets and audit controls.

Ellipsis Review (2026): Cloud Coding Agents and AI Code Review as Code