New in 4.0 — direct the work, Wire runs the commands
Wire Framework
AI-accelerated delivery for data platform engagements
> run what's next dashboard_first · 03-store-dashboards · requirements approved two runnable, no dependency between them conceptual_model-generate — lane · business_rules waived by R-1 mockups-generate — foreground, needs you ✓ 7 entities · validate: PASS 1 decision waiting conceptual_model — approve now, changes, or park?
313 slash commands
A complete command set covering every phase of a data platform engagement, from problem definition through to enablement. A command stops if the step it depends on is not finished — and from 4.0 you can direct the work in plain language instead of typing them.
12 release types
Discovery (diagnostic or modelling-led), dbt development, full platform, platform migration, droughty, and more. Each one carries a machine-readable definition of its own workflow.
Works in Claude Code and Gemini CLI
The same commands run on both runtimes. Install the Wire plugin for Claude Code or the Wire extension for Gemini CLI — specs are shared between them.
New in 4.0
The rules for how a delivery engagement runs used to live inside Wire's own source, as prose. They are now data the framework reads and enforces.
Say what you want, not which command
Wire has 313 commands. You do not have to know them. Say what you want done and Wire works out which command that is, names it before it runs, runs it, and stops where a decision is yours. Typing commands still works, always.
The release director modelAutomatic validation
Generate runs its own validate step when it finishes and folds the PASS or FAIL into its output. Review still requires a passing validate either way.
Meeting transcripts auto-sync
Wire pulls new Fathom call transcripts for the engagement's client into .wire/engagement/calls/ automatically, once per session, with an analytical findings write-up per call.
Agree what the numbers mean first
An optional first step records, per business domain, every competing definition of a metric, the file it came from, what they disagree on, and who approved the decision. Disputed rules generate a reconciliation query that runs immediately rather than surfacing in testing.
Business rules discoveryStart from a model that already exists
Where a client models their data in Modality, Wire reads the entities, sources and relationships from it instead of deriving them again. The requirements are still read, and the difference between the two is raised as a finding.
Reading a Modality modelA machine-readable definition of every release type.
Every release type has a machine-readable definition of its phases and what depends on what. A shared gate reads it and stops a command whose prerequisites are not met. Overriding takes your name and a reason, both recorded.
- A command stops if the step it depends on is not finished.
- Each one carries a machine-readable definition of its own workflow.
- Specs are shared between Claude Code and Gemini CLI.
release_type: full_platform phases: - name: foundation artifacts: - id: requirements - id: conceptual_model depends_on: artifact: requirements step: review status: approved - id: data_model depends_on: artifact: conceptual_model step: review status: approved gate: blocking # name + reason
Works in Claude Code and Gemini CLI.
The same commands run on both runtimes. Install the Wire plugin for Claude Code or the Wire extension for Gemini CLI — specs are shared between them.
Claude Code
Gemini CLI