Switch AI tools without losing your place.

1. SpecWeave CLI

   Open source on npm, MIT. Version 3.1.0, 17 October 2026.

   Writes the spec, the task ledger and the test evidence into your repository, and hands the work to the next tool with specweave handoff. Start here if you work in a terminal with Claude Code, Codex or any agent that reads files.

   npm install -g specweaveCopy

   Then run specweave init in your project.

   How the CLI works

2. SpecWithCode

   Free Mac app. Version 1.0.44.

   A coordinator per project starts worker threads in the agents you already pay for, on your own machines. Each thread shows its SpecWeave increment and task progress, and its reports are saved beside the increment. Start here if you want one window for several agents.

   Download for MacSee SpecWithCode

3. Verified Skill

   Free to browse at verified-skill.com.

   A registry of agent skills. Each shows its source and its security scan, and its evaluation when one exists, so you know what you install before your agents use it.

   Explore skills

1. The spec

   What done means, written once before any agent starts. spec.md

   ## Acceptance Criteria
   - [ ] AC-01: Returning within 24 hours restores the cart and shipping choice
   - [ ] AC-02: A paid order is never restored
   
   ### T-01 Save the cart as a draft
   - AC: AC-01 | Files: src/draft.js | Test: npm test
   
   ### T-02 Restore the draft on return
   - AC: AC-01, AC-02 | Files: src/restore.js | Test: npm test

2. The agents

   Any tool that can read the files claims a task, proves it and moves on.

   1. specweave task claim T-01

      Claude Code takes the first open task. The claim is a line in the ledger, so no other agent takes it too.

   2. specweave task done T-01 --run "npm test"

      Runs the task’s test. Exit 0 is the evidence, and the full output is saved in reports/task-T-01.txt.

   3. specweave task claim T-02

      Claude Code claims T-02 and hits its usage limit halfway through.

      Usage limit

3. The ledger in git

   Every claim, finish and release is one line, committed with the code. ledger.jsonl

   {"t":"T-01","e":"claim","by":"claude@cloud","at":"2026-10-15T14:02:11Z"}
   {"t":"T-01","e":"done","by":"claude@cloud","at":"2026-10-15T14:31:40Z","evidence":"a41c9e2"}
   {"t":"T-02","e":"claim","by":"claude@cloud","at":"2026-10-15T14:32:05Z"}
   {"t":"T-02","e":"release","by":"claude@cloud","at":"2026-10-15T14:58:47Z","note":"handoff"}
   {"t":"*","e":"handoff","by":"claude@cloud","at":"2026-10-15T14:58:47Z","note":"usage limit"}
   {"t":"T-02","e":"claim","by":"codex@laptop","at":"2026-10-15T15:04:19Z"}
   {"t":"T-02","e":"done","by":"codex@laptop","at":"2026-10-15T15:22:51Z","evidence":"npm test → exit 0\nlog: .specweave/increments/0001-resumable-checkout/reports/task-T-02.txt\nHEAD 7d03b1f"}

4. The handoff

   Another tool continues from the same files, not from a chat window.

   claude@cloud hands T-02 to codex@laptop

   1. specweave handoff

      Releases the claim, writes handoff.md with what is next and pushes a snapshot to specweave-handoff and wip/<branch>.

   2. specweave pickup

      In Codex: fetches the handoff, applies the edits and prints T-02 with its two criteria.

   3. specweave task done T-02 --run "npm test"

      Codex finishes T-02 against the same criteria, and the ledger cites the test output.

   4. specweave verify

      Runs test, lint and build, checks every criterion and writes reports/verify.json.

      Pass

Example increment, written in the formats specweave 3.1.0 reads: criteria as - [ ] AC-01: lines, one JSON object per ledger line. See a real pickup in How SpecWeave works.

• See how SpecWithCode runs your agentsSpecWithCode, 2 min 55 s, published 12 October 2026
• Run your first task with CodexSpecWithCode, 1 min 13 s, published 11 October 2026
• Continue a thread on Codex at a usage limitSpecWithCode, 1 min 1 s, published 10 October 2026

All tutorials

• EasyChampLeague and club tools, built across Claude Code threads, two subscriptions and Codex.
• JobWeaveJob search, tailored resumes and interview preparation.
• Verified SkillSkill discovery, source inspection and local evaluations.