Home / Use cases

Local coding agent.

Cursor-style coding. Your hardware. Your code.

Modern coding tools , Cursor, Claude Code, GitHub Copilot Workspace , are extraordinary because they let the model see your whole codebase and act on it. The catch: that means uploading your codebase to someone else's servers. For most professional codebases, that's not acceptable. tailor. ships the same agentic loop, locally.

What the agent can do

Read any file in your project. Run shell commands and see the output. Edit files with surgical patches. Search for symbols across the codebase. Run tests and iterate on failures. Open new conversations branched from the current state. Replay any step to see exactly what the model did and why.

How it works under the hood

Every chat in tailor. is an agent loop. The model picks a tool (read file, write file, run command, search), executes it through tailor.'s sandboxed runner, sees the result, and decides what to do next. The whole transcript , tool calls, outputs, model reasoning , is preserved and replayable, so you can always audit what the agent did.

Connecting your editor

tailor. exposes a drop-in OpenAI-compatible API on localhost:11435. Point Cursor's "Custom Model" setting, Continue's config, your vim AI plugin, or your own scripts at it and they all work transparently. Local model, local context, local execution , but the editor UX you already know.

Which models work well for coding

Qwen 2.5 Coder (7B/14B/32B), DeepSeek Coder V2, Llama 3.1, and Codestral all run well in tailor. and cover most coding workflows. The 32B-class models on a 32 GB machine give you GPT-4-class coding for most tasks. tailor. ships hardware-aware model recommendations so you don't have to guess what'll run.

Questions

How does this compare to Claude Code or Cursor?
Same shape of product (agentic coding loop with tool use), different deployment. Cursor and Claude Code send your code to Anthropic/OpenAI/proprietary servers. tailor. keeps the model on your machine. For sensitive codebases , most professional codebases , that's the difference between using it at work or not.
Can the agent actually run commands on my machine?
Yes, in a sandboxed runner you can configure. By default it can run a curated set of read-only commands (ls, cat, grep) freely and prompts you for anything that writes. You can grant per-project broader access if you trust the loop.
What happens to my prompt history?
It's a local SQLite file. You can search, export, or delete any conversation. It's not synced anywhere by default , peer-to-peer sharing is opt-in per workspace.
Try tailor. free for 7 days.
Full access. No credit card required. Mac, Windows, and Linux.