Roadmap: May-June 2026

[zachlloyd]

We've broken down our roadmap by:

• Warp team committed: stuff in the app that the Warp team itself plans on building this quarter
• Seeking community driver: stuff we would like to work with the community on this quarter

Warp team committed

ACP support (#7326)

• Support Warp as an Agent Client Protocol client so other coding agents can use Warp’s native agent UX
• Supporting ACP will unlock local model support through harnesses that support it (like OpenCode)

High-priority stability and performance fixes

• Warp team has a burndown list of high priority performance, stability and memory fixes that we are working on. Will share the link here shortly.

Improved worktree support + “Project” primitive

• Introduce a “project” primitive in Warp that lets users organize their work around projects
• Worktree navigation: easily switch to a specific worktree. Strawman is a worktree chip + maybe inline menu.
• Worktree management: easy deletion of a worktree when you're done with it. This can be done through the terminal but it's kind of tedious / there's something low-hanging here like a slash command menu to manage known worktrees
• LSP per worktree

Natural language detection improvements

• Improve the classifier we use for triggering our agent to have fewer false positives / negatives

Remote coding support

• Support file-tree, diff-view, LSP and other coding features on remote machines
• Persistent sessions locally and over ssh, pane detaching

Improved third-party agent support

• Continued investment in making Warp the best ADE to build with any coding agent.
• Example features are supporting conversation history, best-of-k, subagents

Seeking community drivers

Tmux control mode

• Use tmux to drive Warp’s native tabs, panes and sessions

Support Warpifying more shells

• Support the native Warp input and block UX within more shells E.g. nushell, elvish

Improved code review

• Side by side diffs
• Selecting code within deleted sections

Warp CLI

• Control all aspects of the warp client app via CLI

Improved keybindings

• Improve our keybinding engine to be more flexible and configurable
• Improve the UI for settings and updating keybindings

Internationalization

• Translate UI text
• Better support for non-English keyboards

Custom prompt chips

• Arbitrary prompt customization

Theming improvements

• Per session / pane / tab / window theming

[Jeffreyone]

三年了国际化终于要来了。

[chan-yuu]

marked

[Koala-MagicSquare]

三年了国际化终于要来了。

But it is Seeking community drivers,这意味着没有确定的时间呀。同样官方把它排到估计是低优先级。好在说总比不说好

[micahnz]

Found myself here via a link in Reddit so I'll give you my number one blocker for rolling out Warp to the rest of my Team.

The current method for securing sensitive outputs/secret redaction falls short of me being able to roll it out for general terminal use within my organization, I've just opened a feature request for this but to re-enforce the point I'll mention it here as well, having AI and Antigenic abilities in the terminal is a huge productivity booster for me, but as it stands I currently have to either copy paste things into another AI client to ensure I'm not sending sensitive context to an LLM or be very careful not to accidentally expose something in an agent session in Warp by running a command that say outputs a Kubernetes secret and I sure as hell can't allow my team members to use Warp and accidentally do something similar, this limits the usefulness of Warp for me and stops me from rolling it out to the rest of my organization where I think it would be genuinely useful.

The current secret redaction is a nice step in the right direction but its too limited to be useful for my operations workflows where I don't just need check for and match for simple patterns like API keys, and currently doesn't support multi-line so I can't even add something that matches say entire Kubernetes secret. What I really need, is to supply a list of commands, or matched command patterns that will redact the entire output of some commands entirely.

For example anything that contains kubectl get secret(s)? would be marked as a secret output and be redacted from LLM context, or say contains "id_rsa", "id_ed25519", "/.ssh/.*" or "^env" to make sure someone doesn't leak a password stored in an environment variable.

This would let me provide a list of commands to be treated as secret to the rest of the team which would let me mitigate some of the risk of using an AI enabled terminal, better yet if I could enforce redaction at an organization level it would go a long way to making this workflow possible for my team and clients.

Along with that, being able to exclude blocks form context manually would also be great once already in an agent session, basically the opposite of attach as agent context so I can filter out any mistakes before continuing on with the session which as far as I understand cannot be done right now.

It would also be useful mark any command as secret before I hit enter or supply some prefix that will run the command and mark it as secret and to be excluded from any contexts including its outputs.

A couple of years ago I would never have entertained the idea of AI in the terminal but I think it is genuinely useful and Warp has an excellent UX that would fit well with my organization, but strong tooling around securing context is probably the biggest thing holding us back and I imagine many others in the industry, you can see it in the comments about anything Warp related anywhere.

Thanks for your consideration.

#10027