secant

Secant runs the workflow. Your agents do the work.

InstallSee how it works
secant · Run Workbench

Your agent works through the step in its own session.

Screens from the Secant Run Workbench, recorded from a scripted run of the Matt Front Spec workflow:
  1. Agent works: Your agent works through the step in its own session.
  2. You decide: The workflow stops at a human gate until you choose how it continues.
  3. Agent asks: When the agent needs a decision, the run waits for your reply.
  4. Step done: When a step ends, the next iteration starts in a fresh session.
  5. Approve: Permission requests from the agent come to you to allow or deny.
  6. Run finished: The run finishes when every step in the workflow is done.

From prompts to a process.

Your coding agents already do capable work. What's missing is the process around them.

By hand

Every handoff still goes through you.

Written once

Write the process down once.

The steps, handoffs, and review points you repeat become a workflow.

Run by Secant

Secant runs the workflow. Your agents do the work.

Each step reaches its agent with the context it needs, and the run stops where you review.

Your workflows

That was one workflow. Build your own.

Any process you repeat can become a workflow.

Your agents

Bring the agents you already use.

Claude Code and Codex stay as they are. Secant coordinates the steps between them.

Keep your agents. Stop carrying the work between them.

Run your first workflow today.

Start with Matt Front Spec, the built-in workflow that turns an idea into production-grade code.

  • Works with your Claude Code or Codex subscription. No API keys.
  • Runs on your machine, in your repo, with your tools.
  • macOS (Apple silicon) · Linux x64 · Windows x64