← All organizations

AI software development and agent orchestration

Augment Code

Augment Code builds AI tools for software teams working with complex codebases. Its Cosmos platform coordinates agents that turn tickets into pull requests, review code, test changes and investigate incidents, with engineers controlling approval checkpoints. Teams configure workflows in natural language and run agents in managed cloud environments or on their own machines. The Auggie CLI brings agent capabilities into existing engineering workflows, while Context Engine MCP supplies codebase context to other compatible coding agents through local and hosted modes.

Founded in 2022 by Igor Ostrovsky and Guy Gur-Ari, Augment Code is led by CEO Matt McClernan. Its technical approach centers on retrieving relevant proprietary code rather than expecting a model to already know it. The Context Engine uses semantic search to select codebase relationships, dependencies and architectural patterns for a task. Exposing that retrieval through MCP lets teams use Augment’s context technology independently of its own agent interface.

Enterprise customers include Intercom and Pearl Technologies. In its customer materials available in 2026, Augment reported that Pearl opened more than 100 pull requests in its first 30 days using Cosmos. In 2024, Augment raised a $227 million Series B at a $977 million post-money valuation, bringing total funding at that time to $252 million.

www.augmentcode.com

5 talks

Newest first

5 speakers at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Start here

  1. Building Self-Coding Agents

    Start here to learn how planning, memory, and tools work together when an agent modifies its own implementation.

    Colin FlahertyAI Engineer Summit 2025

  2. Mentoring the Machine

    See how RFC preparation, incident investigation, Git bisect, and a large gRPC remediation become concrete tasks for parallel agents.

    Eric HouAI Engineer World's Fair 2025

  3. To the moon! Navigating deep context in legacy code with Augment Agent

    Follow a worked example of using an agent to understand unfamiliar assembly and connect that investigation to a simulated lunar landing.

    Forrest Brazeal · Matt BallAI Engineer World's Fair 2025

  4. Vibes won't cut it

    Augment Code's Chris Kelly argues that generating code is not equivalent to engineering production software.

    Chris KellyAI Engineer World's Fair 2025

Messages from the stage

Building agents that can extend themselves

Colin Flaherty connects planning, tool use, memory, and recursive execution to an agent that adds integrations and tests. Graphite workflows and parallel-programming errors expose practical limits.

Delegation still requires engineering ownership

Eric Hou advocates specifying outcomes, context, and constraints while retaining oversight. Chris Kelly connects AI-generated code to continuing responsibilities for architecture, maintenance, debugging, and operations.

Investigating unfamiliar legacy code

Matt Ball and Forrest Brazeal jointly demonstrate codebase investigation through Apollo 11 Guidance Computer assembly. Their session traces a scheduler alarm, distinguishes landing routines, and demonstrates completing missing simulator code.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-27