Magic

Long-context AI models for autonomous, codebase-wide software engineering.

Magic is an AI software engineer platform building long-context coding models — including its LTM-2 model — capable of understanding entire codebases and autonomously implementing complex, multi-file software changes. Many coding AI tools are limited by the amount of context they can hold at once, constraining their ability to reason about large, interconnected codebases; Magic’s research focus on long-context model architecture is specifically aimed at overcoming this limitation, enabling the model to maintain awareness of an entire codebase rather than a limited window around the current task. The platform is positioned exclusively at the enterprise level and differentiates through its research-driven approach to long-context modeling as the core technical bet underlying its product, rather than building on existing foundation models with retrieval-based context management.

Compliance

SOC 2

Visit Magic

Key Features

Long-context architecture: Built around models with substantially extended context windows, specifically engineered to maintain awareness of large, complex codebases in their entirety rather than relying on retrieval-based context management.
LTM-2 proprietary model: Develops its own proprietary long-term memory model architecture, representing a distinct technical approach relative to companies building on existing foundation models.
Autonomous multi-file implementation: Capable of autonomously implementing complex changes that span multiple files, leveraging its codebase-wide understanding to make coordinated, consistent changes.
Codebase-scale understanding: Designed to comprehend the full scope and structure of an entire codebase, supporting more accurate and contextually appropriate code generation for large-scale software projects.
Enterprise-exclusive focus: Targets enterprise customers exclusively, reflecting the scale of codebase and resource investment its long-context approach is designed to address.

Use Cases

For enterprises with large, complex codebases: A large enterprise with a sprawling, interconnected codebase that exceeds the context limits of typical AI coding tools uses Magic to get code generation and editing that accounts for the full scope of the codebase rather than a limited local context.
For organizations implementing complex multi-file changes autonomously: An enterprise engineering team needs to implement a change that spans many files and components uses Magic to autonomously execute the change with awareness of how all the affected parts of the codebase relate to each other.

Pricing MODELS

Enterprise

Pricing Summary

Magic uses a custom enterprise pricing model. No self-serve tiers or publicly listed prices are available — enterprises contact the company directly for a quote.