Backed byY Combinator

Managed Cloud Agents

Run Claude Code and Codex in fully configurable environments without thinking about infrastructure. Watch, steer, and stop every session as if it were on your laptop.

Install on GitHub
Schedule a demo

Free for individuals with a ChatGPT or Claude subscription.

View Full API Reference

Agents-as-a-Service

Start sessions, stream progress, and manage agents programmatically. We handle the cloud environments, permissions, and budgets while you retain full harness control and observability.

Free for individuals with a ChatGPT or Claude subscription.

FOR DEVELOPMENT

Interactive Sessions

Cloud sessions feel like local but they give you an isolated cloud environment with your repositories, dependencies, and tools ready. No more git worktree conflicts.

Learn more about Interactive Sessions
your-org
New sessionK
Dashboard

Today

Yesterday

PS

Outbound webhook deliveries fail permanently on any non-2xx. Add retries with exponential backoff, max 5 attempts, and record each attempt.

Starting sandbox…
0s

Reply to the agent...

Send

FOR AUTOMATIONS

From YAML to API

Define agents in YAML, deploy with a merge, and run them through API calls, schedules, or events.

Learn more about Automations
View Full API Reference
standup-automator.yaml
1
ellipsis:
2
name: standup-automator
3
description: Posts a daily standup digest to Slack
4
trigger:
5
type: cron
6
schedule: "30 8 * * 1-5"
7
session:
8
claude_code:
9
model: claude-opus-4-8
10
prompt: |
11
Each weekday, summarize yesterday's merged PRs, the
12
reviews still waiting, and any failing CI, then post
13
the digest to #engineering and @ the right owners.
14
budget:
15
session: 2
16
17
18
19
20

FOR CODE REVIEW

Custom Code Review

Configure reviews in YAML. Run specialists on relevant changes and add a second pass to filter findings. Ellipsis posts comments directly on GitHub.

Learn more about Custom Code Review
code_review.yaml
1
ellipsis:
2
kind: "code_review"
3
pull_requests:
4
base: ["default"]
5
repositories:
6
include:
7
- "web-repo"
8
- "api-repo"
9
for:
10
bots:
11
include: true
12
exclude:
13
- "dependabot[bot]"
14
review:
15
- name: "correctness"
16
claude_code:
17
model: "claude-fable-5-1"
18
prompt: |
19
Find bugs and regressions in the changed code. Check surrounding code before reporting issues. Skip style, formatting, and speculative findings.
20
- name: "frontend"
21
pull_requests:
22
paths:
23
- "**/*.tsx"
24
- "**/*.jsx"
25
- "**/*.css"
26
claude_code:
27
model: "claude-fable-5-1"
28
prompt: |
29
Install dependencies and start the app. Use Playwright to capture affected pages at desktop and mobile sizes, and inspect the screenshots for visual regressions. Upload screenshots with agent file upload and include the links in your findings.

How it WorksConfigure, control, and observe your agents

  1. environment.yaml
    1
    ellipsis:
    2
    kind: "environment"
    3
    name: "fullstack"
    4
    repositories:
    5
    - name: "web-repo"
    6
    - name: "api-repo"
    7
    variables:
    8
    - name: "NPM_TOKEN" # Stored secret
    9
    compute:
    10
    cpu: 2
    11
    memory: "8GB"
    12
    timeout: "30m"
    13
    hooks:
    14
    before_start: |
    15
    set -e
    16
    cd /sandbox/web-repo
    17
    npm install # Install package.json dependencies
    18
    # Prepare browser tests
    19
    npx playwright install chromium
    20
    npm run build
    21
    cd /sandbox/api-repo
    22
    npm install # Install package.json dependencies

    Define an Environment

    Define repositories, dependencies, environment variables, and startup commands in YAML, then reuse the environment across sessions and automations.

    Environments
  2. automation.yaml
    1
    ellipsis:
    2
    name: "issue-investigator"
    3
    session:
    4
    claude_code:
    5
    model: "claude-fable-5-1"
    6
    prompt: "Investigate the issue and report your findings."
    7
    environment: "fullstack"
    8
    permissions:
    9
    github:
    10
    repositories:
    11
    - "web-repo"
    12
    - "api-repo"
    13
    permissions:
    −
    contents: write
    14
    contents: read
    −
    pull_requests: write
    15
    pull_requests: read
    16
    17
    18

    Set Permissions Scope

    Choose which repositories a session can access and restrict its GitHub credentials to read-only when it only needs to inspect code.

    Permissions
  3. create_automation.py
    1
    import os
    2
    from ellipsis import Ellipsis, models
    3
    4
    client = Ellipsis(api_key=os.environ["ELLIPSIS_API_TOKEN"])
    5
    6
    request = models.CreateAutomationRequest.model_validate({
    7
    "automation": {
    8
    "ellipsis": {"name": "check-tests"},
    9
    "session": {
    10
    "claude_code": {
    11
    "prompt": "Run the tests and report failures."
    12
    },
    13
    "environment": "api-environment"
    14
    }
    15
    }
    16
    })
    17
    18
    client.automations.create(request.automation)

    Spawn from Anywhere

    Spawn sessions through the API, CLI, web app, or directly from Slack, GitHub, Linear, and Sentry.

    Automations
  4. Steps (1)
    1thinkingPlan the fix for the connection leak
    Budget$0.03 / $1.00
    Tokens2.0K

    Observe the Session

    Follow sessions live and inspect each step, tool call, and result.

    Sessions
  5. session-record.json
    1
    {
    2
    "kind": "claude_code",
    3
    "source": "claude_code",
    4
    "record_format": "claude_jsonl@1",
    5
    "record_type": "assistant",
    6
    "session_id": "session_2f8c1d",
    7
    "turn_id": "turn_9a41",
    8
    "feed_seq": 42,
    9
    "payload": {
    10
    "type": "assistant",
    11
    "message": {
    12
    "role": "assistant",
    13
    "model": "claude-fable-5-1",
    14
    "content": [
    15
    { "type": "text", "text": "Running the pool test suite." },
    16
    {
    17
    "type": "tool_use",
    18
    "name": "Bash",
    19
    "input": { "command": "pytest -k pool" }
    20
    }
    21
    ],
    22
    "stop_reason": "tool_use"
    23
    },
    24
    "uuid": "22222222-2222-4222-8222-222222222222"
    25
    },
    26
    "tools": ["Bash"],
    27
    "cost": 0.0142,
    28
    "created_at": "2026-09-10T14:00:00Z"
    29
    }

    Download the exact agent output

    Inspect unmodified Claude Code or Codex output and download the session logs as a compressed archive.

    Observability
  6. nightly-migration-checker
    Cost
    $0.60
    Tokens
    12,000
    Session
    12%
    Day
    68%
    Week
    40%
    Month
    28%
    Running

    Keep agent spending under control

    Set session spending limits and track costs by developer or automation.

    Budgets

  7. Observe your Fleet at Scale

    Track sessions, costs, token usage, and tool activity across your team from one dashboard.

    Observability

Secure, Managed Infrastructure
that is very easy to use

Ellipsis is built with governance at scale in mind. Permissions are defined before a session starts. Every agent action is logged. You have full control over your data longevity.

More about security
SOC 2Type ICertified
SOC 2Type IIIn progress

Scoped credentials

Minted per session, scoped to the repos you allowed, revoked with the sandbox.

Zero source retention

Your code lives only in the session sandbox and is deleted with it.

Full audit trail

Every command, file change, and tool call an agent made, kept after the sandbox is gone.

Frequently asked questions