The best AI coding tools and new workflows
8 picks · By SaveToList Picks ·
Current agents, recent integrations and meaningful releases, chosen for distinct coding jobs. New launches are labelled separately from established picks. Curated by SaveToList Picks. Follow meaningful releases and workflow discoveries, with setup and access notes. Recent creator and community sources are linked on the picks they informed. Dates distinguish launches, meaningful updates and renewed recommendations; established foundations are kept where they still earn their place.
Follow this list in SaveToList to keep future additions in reach.
Follow the launch integration to connect Flint to a repository-working agent. A timely experiment for developers who want model choice and control over where code runs. October 21 launch integration for Recursive Flint. Open-source agent and downloadable weights; hardware and any cloud service still cost money.
Visit source
Search prior agent sessions and retain reviewed lessons across tools, so the next debugging session can find what already worked. Most useful when you actually switch between agents. October 20 maker update: shared coding-agent history and reusable skills. MIT-licensed software; local history is available. Hosted forwarding has separate terms. Review what history is collected before enabling hosted forwarding.
Visit source
Use the current app, CLI or IDE to work through a contained project. Follow the new skills migration guide when carrying over an older workflow rather than copying obsolete setup advice. October 2026 release plus an October 19 workflow migration. Plan, model and usage limits vary; check the current plan before relying on a free tier. Availability may still be rolling out.
Visit source
Use an open-source coding agent while choosing the model provider that fits your project. A practical option when you want to keep the workflow but change hosted or local models. OpenCode 2.0: October 12 worktrees make parallel repository work more concrete. Some free models are offered; other providers can charge for inference.
Visit source
Use codebase chat, inline suggestions, and coordinated edits in one editor. Pick this if you want to steer an assistant while reading and changing the surrounding code yourself. Cursor’s October 16 update adds Cloud Agent Canvases and environment improvements. Commercial editor; model availability and usage depend on the plan.
Visit source
Start with one reproducible bug and ask Claude Code to explain the relevant files, make a patch, and run the checks. Best when you already have a repository and can review changes. Eligible Claude plans or metered Console access; usage limits apply.
Visit source
Delegate a feature, refactor, or code review, then inspect the proposed changes. Useful for a contained task you can describe with clear acceptance criteria; local and cloud workflows serve different jobs. Access and usage limits depend on your ChatGPT plan and chosen workflow.
Visit source
Add assistance to an IDE you already use, or connect a GitHub issue to an agent task. A useful fit when review and collaboration already happen in GitHub pull requests. Free and paid Copilot plans have different features and usage allowances.
Visit source
Turn on JavaScript to follow this list or save a copy.
