llms.txt
Compare with other sites' llms.txt files →# MEGA
> The shift happened. The only question is how deep you go.
Canonical home: https://mega.dev/
MEGA can run bounded, purpose-specific model operations exposed by its API, but it does not run caller-selected code, shell commands, filesystem actions, or unrestricted network requests. Every API and MCP action is limited by the permissions of the authenticated MEGA identity.
## When to use MEGA
- [MEGA program](https://mega.dev/): Use when a user asks what MEGA teaches, who the program is for, or how its four-week course works.
- [MEGA Drops](https://mega.dev/drops): Use when a user asks for MEGA field notes, articles, or downloadable artifacts about building with agents.
- [MEGA developer resources](https://mega.dev/developers): Use when an authorized MEGA user asks an agent to inspect or perform a supported MEGA operation.
## How to call MEGA
- [Canonical operator OpenAPI](https://mega.dev/api/operator/v1/openapi.json): For direct HTTP clients; operation IDs, request schemas, responses, and authorization requirements are authoritative.
- [Remote MCP discovery](https://mega.dev/.well-known/oauth-protected-resource/mcp): For MCP clients; connect to https://mega.dev/mcp, complete OAuth, then use operations.search, operations.describe, and operations.invoke.
- [Drop authoring brief](https://mega.dev/labs/drops/agent-brief.md): For authorized agents drafting a MEGA Drop; follow its validate, create, update, review, and owner-directed publish contract.
## Public pages
- [It’s time | MEGA](https://mega.dev/): The shift happened. The only question is how deep you go.
- [MEGA Developer Resources](https://mega.dev/developers): MEGA developer resources for the operator API, OpenAPI, OAuth, CLI, and MCP.
- [Drops | MEGA](https://mega.dev/drops): Field notes from building with agents — every Drop ships something you can run.
## Posts
- [Accessible web for humans and agents](https://mega.dev/accessible-web): We've made a better web for humans. Semantic HTML, alt text, and captions help agents read and act, too. Treat agents as a new kind of user.
- [AI Changed Programming. But How?](https://mega.dev/ai-changed-programming): AI changed more than how we write code. Packages, cloud agents, durable context, and bounded autonomy shift programming toward shaping reliable environments.
- [Towards Autonomous Product Development](https://mega.dev/autonomous-product-development): Language models got smarter, but they still lack our context. Shape the environment so agents can work without us: always on, in the cloud, with real limits.
- [From Visual Editors to an Agentic-First Web](https://mega.dev/from-visual-editors-to-agentic-first-web): What moving 924 pages from Webflow to Astro unlocked: an agentic-first website, connected documentation, and publishing workflows that go beyond the visual editor.
- [Model pricing isn't reliable anymore](https://mega.dev/model-pricing): Model families keep multiplying and provider prices keep changing. Choosing the right one for a real task takes more than a single glance at the price tag.
- [Don't make the agent remember how to do its job](https://mega.dev/pi-kody-integration): How Pi and Kody turn exploratory agent work into a tested package with remote runtime, durable state, and a schedule that keeps running after the laptop closes.
- [Your reasoning effort level is too high](https://mega.dev/reasoning-effort): Every model now comes with a reasoning dial. Higher settings give the model more room to think, but choosing the right level is less obvious than it looks.
- [Agents shouldn't live on your laptop anymore](https://mega.dev/remote-agent-setup): Build a remote home for coding agents with an always-on VPS, Tailscale, Herdr, Pi, and Moshi, then control the same sessions from any device.
- [Rethinking AI with Jev](https://mega.dev/rethinking-ai-with-jev): John Lindquist's October 8 MEGA livestream on Jev: a fast, cheap decision model that picks from a menu you wrote and returns a score your code can act on.
- [How Kent Ships Features Without Reading the Code](https://mega.dev/ship-without-reading-the-code): Kent C. Dodds' real delivery loop on his own SaaS: fifteen principles for shipping agent-written code you haven't read, with the full session recording.
- [We're system designers now](https://mega.dev/system-designers-now): Angie Jones on the MEGA podcast: typing code may be solved, and the job that remains is designing the system agents can run without you in the middle.
- [It's time: the work changed](https://mega.dev/the-work-changed): The four short opening tutorials from MEGA's August 25 live: Angie, John, Kent, and Theo on autonomy, images, primitives, and prompting the whole outcome.
- [Whole products can now be prompts](https://mega.dev/whole-products-as-prompts): Theo Browne on the MEGA podcast: features and whole products can now be prompts or markdown files, and the engineering that still has to stay software.
- [I was wrong about MCPs](https://mega.dev/wrong-about-mcps): I treated MCPs as service connectors. The useful shift is a capability layer: search, execute, and reusable primitives an agent can compose on its own.
## Challenge
- [MEGA challenge](https://mega.dev/challenge.md): Consent-gated, agent-run assessment instructions.