Ellipsis
Also known as: Ellipsis, ellipsis.dev, Ellipsis.dev
AI code review platform that reviews every GitHub pull request for bugs and style violations, generates tested fixes, and lets teams deploy configurable coding agents in the cloud.
Ellipsis is an AI code review and coding agent platform that installs as a GitHub app and reviews every pull request automatically. When a pull request opens or updates, it scans the diff for logical bugs, null pointer and conditional errors, anti patterns, and style guide violations, then comments inline on the exact lines that matter and generates a summary of the change. Teams write their style guide in plain English and Ellipsis flags violations against it, and the reviewer learns over time which comments a team values so feedback grows more relevant. Founded in 2024 by Nicholas Moy and James Murdza and backed by Y Combinator in the W24 batch, Ellipsis reports accelerating merges by roughly thirteen percent and is used by teams including PromptLayer and Warp.
Beyond review, Ellipsis acts as an AI teammate you assign work to from a GitHub comment. Tag it on a bug report or a feature request and it produces working, tested code, executing what it writes in a sandbox the way a developer would rather than only suggesting edits. It makes changes across multiple files from natural language requirements, and it will never commit without explicit permission, optionally raising a side pull request so its changes stay clear. Developers can also tag it with a question to get answers grounded in the codebase, which helps with onboarding and triage, and it can generate weekly digests that summarize what changed and why it matters.
In 2026 Ellipsis extended into a platform for deploying, managing, and observing coding agents in the cloud. Each agent is defined in a single YAML file that sets its prompt, its model, its trigger, the tools and MCP servers it can use, its permission boundaries, and a spend cap per run. Committing the file deploys the agent, a pull request changes it, and a revert rolls it back, so agents are versioned and reviewed like code. Example agents keep documentation in sync, triage failing CI, guard migrations, and automate standups, each running in its own isolated sandbox with a trace that is visible to the whole team. It connects GitHub, Slack, and Linear, works with models including Claude, Codex, and Gemini, and reports use by more than four hundred engineering teams.
Ellipsis is SOC 2 Type 1 certified and does not persist source code, holding it in a private AWS VPC only while a job runs and clearing it afterward, with least privilege access and per run permission and spend controls. Review is GitHub centric with no IDE, editor, or command line surface, and the standard delivery is a hosted cloud service, though the team will set up a private instance in a customer's own AWS, GCP, or Azure account on an enterprise basis. Pricing is simple: it is free with no feature limits for public GitHub repositories, and private repositories cost twenty dollars per developer each month for unlimited use across all repositories, billed as a fixed number of assignable seats, with a seven day free trial and no credit card required.
Vendor details
Canonical URL
https://www.ellipsis.dev
Category
Coding agent
Subcategory
AI code review and coding agent platform
Funding status
Independent. Founded in 2024 by Nicholas Moy and James Murdza and backed by Y Combinator in the W24 batch, Ellipsis raised roughly two million dollars in seed funding. It reports use across more than twenty nine thousand codebases and by more than four hundred engineering teams, with named users including PromptLayer, Warp, and Onfleet.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Installs as a GitHub app, with GitHub as its primary and review surface, and accepts tasks by tagging it in Slack, GitHub, or Linear. Its cloud agent platform lets each agent connect tools and MCP servers under scoped permissions. Ellipsis acts by commenting inline on pull requests, generating tested code, opening pull requests, answering codebase questions, logging issues, and producing weekly change digests. Review is GitHub centric with no IDE, editor, or command line extension.
In practice
Your team ships fast and pull request review has become the bottleneck. Ellipsis reviews every pull request in GitHub, flags real bugs and style violations on the exact lines, and summarizes the change.
You maintain a busy open source project and cannot review every contribution yourself. Ellipsis is free with no feature limits on public GitHub repositories, so it reviews contributor pull requests and answers questions for you.
You want routine engineering chores handled automatically. On Ellipsis you define each agent in one YAML file with its model, trigger, and spend cap, deploy it by commit, and watch its trace as a team.
Sources & related URLs
Related / legacy domains
Capability coverage
7.0 / 14 capabilities · 50%
| Integrations & Tool CallingConnects GitHub, Slack, and Linear and acts by commenting inline on pull requests, generating tested code, opening pull requests, answering questions, and logging issues, with per agent tools including MCP servers, a real integration and tool calling surface that stays concentrated on the GitHub centered developer workflow with no IDE or command line surface. | Partial |
|---|---|
| Workflow OrchestrationRuns a multi agent review pipeline of parallel comment generators and a filtering stage with a shared code search subagent, and its cloud platform deploys and manages many independently triggered agents that run autonomously in their own sandboxes, strong orchestration that centers on advisory review and fleets of scoped agents rather than one deep autonomous end to end pipeline. | Partial |
| Knowledge Grounding & RAGGrounds review, code generation, and codebase chat in a code search subagent that combines keyword and vector search with tree-sitter AST parsing and pull request indexing, real retrieval grounding that supports the workflow rather than being sold as a deep whole codebase index headline. | Partial |
| Human Oversight & GuardrailsAdvises through inline review comments while humans merge, never commits code without explicit permission, can raise a side pull request to keep its changes clear, and enforces per run permission boundaries on platform agents, a strong human oversight surface that gates and flags rather than running a hard blocking runtime enforcement engine. | Partial |
| Security, Identity & GovernanceIs SOC 2 Type 1 certified, retains zero source code by holding it in a private AWS VPC only during a job, and applies least privilege access by default with per run permission and spend controls, a solid security posture, though a full identity and governance matrix with SSO, SAML, SCIM, RBAC, and audit is not enumerated. | Partial |
| Observability & AuditabilityHeadlines observability by letting a team deploy, manage, and observe its agents in the cloud with each agent trace visible to the whole team, plus review visibility through pull request summaries and weekly change digests, real agent trace and review observability short of a comprehensive tracing and analytics suite. | Partial |
| Memory & State PersistenceLearns over time which review comments a team values and adapts its feedback, but this is feedback loop adaptation grounded in the codebase and review history, counted under knowledge grounding, not a distinct persistent agent memory product. | Unable to verify |
| Deployment & Data ResidencyShips as a hosted cloud service by default, but the team will set up a private instance in a customer's own AWS, GCP, or Azure account on request, a private instance deployment option beyond pure SaaS, though short of a self serve on premises or air gapped self host. | Partial |
| Prebuilt Agents, Templates & PacksShips example agents such as a standup automator, a migration guard, an on call assistant, and a CI failure triager as copyable YAML templates that teams adapt and deploy, reusable agent scaffolding, though closer to starter templates than a large library of prebuilt agents. | Partial |
| Triggers & Channel CoverageTriggers automatically on pull request open and update, supports scheduled and polling triggers defined per agent in YAML, and accepts tags across GitHub, Slack, and Linear, broad trigger and channel coverage within a GitHub centered envelope with no IDE or command line surface. | Partial |
| Model Flexibility & RoutingLets each platform agent choose its model from supported options including Claude, Codex, and Gemini set in the agent YAML, and internally mixes multiple models across review subagents, real model choice, though it does not support bring your own model API keys or a multi provider routing gateway. | Partial |
| APIs, SDKs & MCP ExtensibilityExposes an agents as code model in which each agent is a versioned YAML file that can attach tools and MCP servers under scoped permissions, a genuine programmatic extensibility surface through MCP and declarative config, though without a documented public platform SDK. | Partial |
| Testing, Debugging & OptimizationMakes bug and quality detection the core product, reviewing every commit for logical errors, anti patterns, style guide violations, and basic security issues, and generating working tested code that it executes in a sandbox, a headline testing, debugging, and quality capability. | Full |
| Browser & Computer UseRuns every agent in its own isolated cloud sandbox to execute and test code, and equips agents with browser tools such as opening a preview URL and capturing a screenshot, real sandbox and computer use, though not general autonomous browser automation as the core product. | Partial |
Pricing
Free (public repos, no feature limits) · Developer $20/dev/mo (private repos, unlimited use across all repos) · 7 day free trial
Per developer per month subscription. Buy a fixed number of seats, assign them to developers, and each assigned seat gets unlimited use across all repositories. Free and unlimited for public repositories. Seats can be added, removed, or reassigned at any time.
Included quota
Free: unlimited on public GitHub repositories with no feature limits, install to use. Developer ($20/dev/mo): private repositories, AI reviews on every commit, pull request summaries, unlimited code generation, codebase question and answer, style guide enforcement, and unlimited use across all repos. Seats are purchased in a fixed count per month and assigned to specific developers; unlicensed developers are ignored. 7 day free trial, no credit card. Enterprise adds a private instance in the customer cloud on request.
What is public
Both tiers are public and exact: free for public repositories and twenty dollars per developer per month for private repositories. Enterprise terms, including the private instance option, are contact us and not publicly priced.
Billing mechanics
Per developer per month seats purchased in a fixed count and assigned to developers; each seat is unlimited across all repositories. Free and unlimited for public repos. 7 day trial, no credit card. Seats can be added, removed, or reassigned at any time.
Cost watchouts
At twenty dollars per developer, a thirty person team is six hundred dollars a month, so seat count is the main cost driver and seats should be assigned deliberately. Platform agent runs consume compute bounded by the per run spend cap you set in each agent's YAML.
Variable cost rationale
Pricing is a flat twenty dollars per developer per month with unlimited use across all repositories and no metering of reviews, generations, or tokens, so spend scales only with the number of developer seats and is fully predictable. The cloud agent platform adds a per run spend cap on each agent, which bounds agent execution cost rather than exposing open ended usage.
Additional watchouts
Review is GitHub only with no IDE, editor, or command line surface. The standard product is a hosted cloud service with no self serve self host; a private instance in the customer's own AWS, GCP, or Azure account is available only as an enterprise contact us option. Requires admin level GitHub access, which can stall in enterprise IT approval. Review depth on complex cross file logic and on niche languages such as Go and Rust is weaker than on mainstream languages. No bring your own model API keys.
Overage / add-ons
No metered overage. Each purchased seat includes unlimited use of the product across all repositories, so cost scales only with the number of developer seats. Platform agents are additionally bounded by a per run spend cap set in each agent's YAML.
Sales call required
No — self-serve available
Free / trial
Free and unlimited for public GitHub repositories with no feature limits (install to use). Private repositories: 7 day free trial with no credit card required; during the trial all developers in the GitHub organization have access.
Lowest paid plan
Developer plan at twenty dollars per developer per month for private repositories: AI code reviews on every commit, pull request summaries, unlimited code generation, and codebase question and answer, with unlimited use across all repositories.
Commercial notes
Positioned as a low cost, self serve code review and coding agent tool at a flat per seat price with no usage metering, undercutting reviewers that bill by lines or usage. Value is highest for GitHub first teams that want review plus lightweight autonomous agents without procurement. The free public repo tier is a strong pull for open source maintainers.
Key ambiguities
Support beyond GitHub is reported inconsistently across third party sources; GitHub is the confirmed primary and review platform. Enterprise and private instance pricing are not surfaced.
Cancellation / refund
Monthly per developer seats with a 7 day free trial and no credit card to start. Seats can be added, removed, or reassigned at any time and the plan can be cancelled at any time. Specific refund terms are not detailed.
Support SLA / resale
SOC 2 Type 1 certified; report available on request by emailing the team or scheduling a call with a founder. Code is processed in a private AWS VPC and not persisted. No reseller or white label program surfaced.
Missing data
Enterprise pricing and the cost of a private instance in the customer cloud are not public. Volume or annual discounts, if any, are not documented. The commercial boundary between the per seat product and paid platform agent runs is not fully detailed.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- Aider — Open-source, model-agnostic terminal coding agent that works…
- Anthropic Claude Code — Anthropic's terminal and IDE-native coding agent with sub-agent…
- Augment Code — AI coding assistant evolving into an AI-native engineering platform
- Blackbox AI — AI coding assistant for code generation and developer workflows
- Blink — Agentic AI app builder that turns a natural language prompt into a…