Workspai.dev

Release Notes

Authoritative Workspai changelogs, releases, compatibility history, and rules for interpreting shipped capabilities.

This page points to the authoritative release history for Workspai CLI and the VS Code surface. Use those primary sources when checking whether behavior is shipped, compatible, changed, or still planned.

When reading a release, distinguish:

  • shipped CLI behavior,
  • synced contract changes,
  • VS Code surface compatibility,
  • docs updates,
  • planned roadmap items.

Current CLI baseline

The current CLI baseline is Workspai CLI 0.72.1. It provides runtime-truth repair eligibility, workspace-owned polyglot lifecycle planning, typed Windows CLI-resolution evidence, and script-free npm 11 installation. The current baseline also includes project-owned Graphs, evidence-qualified Skills, Live activity projections, measurable retrieval, Proof-Carrying Change, and enterprise native-kit baselines introduced across the preceding compatible releases.

The website mirrors the complete 0.72.1 contract catalog of 123 contracts. The extension's minimum compatible CLI floor remains a separate contract and can be lower than the current release; a backward-compatible CLI release does not automatically raise that floor.

Evolution since 0.52

ReleaseShipped boundary
0.52Verified Engineering Goals and evidence-backed definitions of done
0.53CLI-owned Workspace Repair Engine with immutable plans, approval, checkpoints, verification, and rollback
0.54Typed multi-ecosystem dependency materialization and target-aware repair receipts
0.55Approval bound to causal actions, typed decision causes, and multi-project repair scope
0.56Universal Doctor diagnosis, live graph freshness, C/C++ and polyglot runtime depth, scoped retrieval, and portable external proofs
0.57Governed evaluation reports and bounded Graph retrieval efficiency evidence
0.58Natural-language Goal entry, durable lifecycle state, and agent handoff artifacts
0.59Canonical project agent entry and portable grounding across supported model hosts
0.60Context-aware Goal scope selection and canonical runtime binding
0.61Enriched agent context, dynamic project-aware grounding, and repair replanning
0.62Bounded-first agent entry and runtime-aware operational Skills
0.63Sequential causal repair queues and portable linked-project transactions
0.64Contracted Create Planner semantics, language-aware imports, and platform-aware setup recovery
0.65Project taxonomy, governance and test surfaces, proof-bounded bindings, incremental Graph reuse, and repair qualification
0.66Project-owned Graph projections, evidence-qualified operational Skills, adaptive project context, and MCP interoperability
0.67Live activity Board/fleet contracts and a provenance-aware bounded-retrieval benchmark
0.68Revision-bound Live evidence references and stronger governed agent-repair recovery
0.69Proof-Carrying Change, digest-linked Decisions, predicted-versus-actual Graph deltas, and portable capsule validation
0.70Deletion tombstones, overlapping physical-artifact reconciliation, and stronger nested-project qualification
0.71Stable Goal/PCC startup across operational telemetry, observable cold initialization, and enterprise native-kit baselines
0.72Runtime-truth repair prerequisites, workspace-owner lifecycle planning, and large-repository retrieval improvements
0.72.1Script-free npm installation and typed Windows CLI-resolution Doctor evidence

Changing a manifest is still progress, not completion. A governed repair must reconcile its materialized dependency state, run its declared checks, refresh the exact evidence producer, and return to canonical verification. The CLI—not an IDE or model—owns transaction closure and rollback.

Primary sources

How to interpret claims

A capability is shipped when its owning surface provides the corresponding command, contract, report, or user-facing behavior. Roadmap language and RFCs describe direction; they are not release evidence.

On this page