Aftergradient is a practical reference for AI coding agents and the people who work with them.
The name comes from model training. Gradient descent is how a model learns; what comes after the gradient is the part where people put the trained model to work. That is the part this site covers: the terminal, the editor, the config files and the bill.
What you’ll find here
- Guides for Claude Code, OpenAI Codex, Cursor, GitHub Copilot, and Gemini CLI and Google Antigravity. Each one answers a single job: set this up, fix this, which plan do I need, which tool fits this work.
- Plans tracker: every coding-agent plan each vendor lists, with price, usage wording and the date we last read the vendor’s page.
- MCP directory: MCP servers with install snippets for each client, copied from the server’s own docs.
- Skills directory: agent skills and skill collections with their source, license and install command.
- Updates: what changed in each release, from official changelogs.
How the work is done
Commands, flags, file paths, prices and limits come from the vendor’s own documentation, pricing page, changelog or repository, and the source is linked next to the claim. If a vendor doesn’t publish something, we leave it out rather than guess. Every guide says which version or docs date it reflects and when it was last checked.
We don’t claim to have tested anything we haven’t. Where a guide describes a command, it describes what the documentation says it does. Setups shared by developers are credited to the person who shared them and linked to the original.
The full rules are in our editorial policy.
Who writes it
Guides are published by the Aftergradient desk under the byline Aftergradient Editors. Aftergradient is independent: it is not affiliated with Anthropic, OpenAI, Anysphere, GitHub, Microsoft or Google, and product names belong to their owners.
Get in touch
Found an error, or a release we missed? Use the contact form. For sponsorship, see Advertise.