Ellipsis

Ellipsis is a cloud platform designed to run managed AI coding agents in isolated sandboxes. It allows teams to define their agents using YAML configuration files directly inside their code repositories. The platform aims to help engineering organizations automate tasks like reviewing pull requests, building features, and investigating production issues.

• Run managed AI coding agents in short-lived, isolated cloud sandboxes with limited credentials
• Define agents as code using YAML configuration files directly in the repository
• Enforce spend limits per session, per developer, per day, or per month
• Maintain full observability logs including thinking steps and tool calls
• Automate code reviews, bug fixes, and feature building triggered by cron schedules, pull requests, or CLI commands
• Support multiple models including Claude, Codex, and Gemini through configurable harnesses

Pros

• Provides complete isolation for parallel agent execution so multiple agents on the same repository do not interfere with each other
• Ensures tight governance by minting scoped credentials per session instead of using human developer credentials
• Offers flexible pricing without per-seat fees, allowing an entire team to access the platform
• Tracks detailed transcripts and logs that outlive the temporary sandboxes for complete audit trails

Cons

• Requires managing and writing YAML configuration files and agent prompts within the repository
• Pricing depends heavily on token consumption, CPU hours, memory hours, and platform fees which can scale with complex tasks
• Open-source usage is not universally free and requires a case-by-case evaluation

Ellipsis fits engineering teams and platform developers looking to automate repetitive maintenance tasks, code reviews, and issue triage through cloud-hosted coding agents. It is well-suited for organizations that want strict infrastructure control, budget caps, and audit logs for AI actions. It is a poor fit for teams that prefer not to define agents via repository YAML files or those seeking a completely free tool without usage-based metering.

Ellipsis uses a usage-based pricing model with no per-seat fees or idle charges. New organizations receive a one-time $100 credit, while individual accounts receive a $10 credit. The pricing breaks down into token usage at cost with no markup, CPU billed at $0.142 per vCPU-hour, memory billed at $0.024 per GB-hour, and a platform fee equal to 100% of the token usage. The platform offers three options: Ellipsis Cloud (self-serve), BYOK (bring your own model keys via AWS Bedrock or Anthropic), and Private Cloud (self-hosted in an AWS VPC with platform fees only).

Unlike local coding agents or tools that run directly on developer credentials and machines, Ellipsis acts as a managed infrastructure platform for agents. It handles sandboxes, credential scoping, budget caps, and logs centrally. Rather than locking users into a single model, it allows teams to swap out harnesses and models while keeping the underlying infrastructure consistent.

Ellipsis integrates with GitHub, Claude, Codex, Gemini, AWS Bedrock, and Anthropic.

Ellipsis is worth trying for engineering teams seeking a governed cloud environment to run automated coding agents with detailed observability and strict budget caps. The self-serve usage-based model allows teams to test the platform using initial credits without committing to per-seat licenses. However, teams that prefer zero platform fees or lack the capacity to configure YAML-based agent definitions may find the setup process less suitable.