What Is Ellipsis and What Does Its "Agents as a Service" Platform Do?

Ellipsis is a managed cloud platform for running coding agents. You define an agent as a config file in your repository, and Ellipsis runs it in the cloud to build features, fix bugs, review code, and investigate production issues. It fits teams that want automated coding work tied to their own repos but don't want to build and operate the agent infrastructure themselves.

The core idea: agents as config files

Instead of clicking through a UI to set up an assistant, you commit a YAML file to your repo. Ellipsis reads that file and runs the agent according to it. The platform's own framing is "Agents as a Service": you bring the agent definition, Ellipsis handles the infrastructure, permissions, logging, and budgets.

A minimal agent definition from the documentation looks like this:

ellipsis:
  name: standup-automator
  description: Posts a daily standup digest to Slack
  trigger:
    type: cron
    schedule: "30 8 * * 1-5"
  session:
    claude_code:
      model: claude-opus-4-8
      prompt: |
        Each weekday, summarize yesterday's merged PRs, the reviews still
        waiting, and any failing CI, then post the digest to #engineering
        and @ the right owners.
  budget:
    session: 2

The parts you control are visible in the file: the trigger (here a cron schedule), the session harness and model, the prompt, and a session budget. Because the definition lives in the repo, it's versioned alongside the code it acts on.

What the agents actually do

The platform lists three broad use cases:

• Feature generation — agents that build features from a prompt or trigger.
• Background tasks — scheduled or event-driven jobs, like the standup digest above.
• Code review — agents that react to pull request activity.

The documentation's examples show the range. A migration-guard agent and an oncall-assistant agent appear alongside the standup automator. Another example, check-api-changes, triggers on a pull request push and checks changed API routes for missing test coverage:

ellipsis:
  kind: agent
  name: check-api-changes
  trigger:
    type: react
    pull_request:
      on: [pushed]
    repositories: [api-repo]
  session:
    claude_code:
      model: claude-opus-5-5
      prompt: |
        Check the changed API routes for
        missing test coverage.
  environment:
    repositories:
      - name: api-repo
    hooks:
      after_checkout: |
        cd api-...

Note the hooks section: you can run setup commands after checkout, so the agent's environment is prepared the way your project needs.

Any agent, any model

A distinguishing feature is that the harness and model are swappable. The docs describe changing one line to move from Claude Code to Codex, or from Anthropic to OpenAI, without rebuilding the agent. The example diff replaces claude_code with codex and claude-opus-5-5 with gpt-6-sol while keeping the same prompt and environment.

This matters if you want to compare models on the same task, or if your organization standardizes on a different provider later. The agent logic — trigger, prompt, repositories, hooks — stays put.

How you get started

Two entry points are documented:

• Install on GitHub — connect the platform to your repositories.
• API — the docs list endpoints including POST /v1/agents (create an agent), POST /v1/agents/{id}/sessions (start a session), and POST /v1/sessions/{id}/messages (message a session).

There is also a full YAML reference and a documentation site, plus a "Schedule a demo" option. The site shows a Pricing page and a sign-in flow, but the provided material doesn't state prices or whether any tier is free — check the pricing page directly before assuming cost.

What Ellipsis handles for you

The platform's stated responsibilities are the operational layer around the agent:

• Infrastructure — the compute the agent runs on.
• Permissions — access to your repositories and services.
• Logging — a record of what agents did.
• Budgets — per-session limits, as shown by the budget: session: 2 field.

That division is the practical difference from a plain AI coding assistant in your editor. An editor assistant helps a human who is present and driving. Ellipsis runs agents on triggers — schedules, pull request events — without someone watching each step, which is why permissions, logging, and budgets are part of the product rather than afterthoughts.

When this fits and when it doesn't

Ellipsis is a reasonable fit if you want recurring or event-driven coding work (digests, review checks, issue triage) automated against your own repos, and you'd rather write a YAML file than maintain agent runners. It's less obviously a fit if you only want in-editor suggestions while you code, or if you need to confirm exact pricing and limits first — those details aren't in the material here, so verify them on the pricing page and in the documentation before committing.