Skip to main content

AxonFlow v7.1.1 Release Notes

AxonFlow v7.1.1 is the patch release that made the v7.1.0 feature set behave consistently in real install-and-use scenarios.

The short version:

  • v7.1.0 introduced explainability, session overrides, richer block context, and checkpoint-aware workflow recovery
  • live E2E against the released plugins found gaps that synthetic tests had missed
  • v7.1.1 closes those gaps so the same features behave consistently across HTTP plugin clients, MCP tool calls, workflow-control paths, and audit search

This release matters most if you are using OpenClaw, Claude Code, Cursor, or Codex with the new explainability and override workflows.

What v7.1.1 fixed on the platform side​

The patch closes the parity gaps surfaced by post-release E2E:

Richer block context now survives both plugin paths​

The initial rollout covered the new context shape conceptually, but not every plugin path emitted it correctly. v7.1.1 makes sure the platform returns and propagates richer decision context on both:

  • the direct HTTP MCP check path used by the OpenClaw client
  • the MCP tools/call path used by Claude Code, Cursor, and Codex

That means decision_id, risk_level, matched-policy context, and override hints now appear consistently where the plugin expects them.

Explainability works for MCP-path decisions​

Early release testing found that some MCP-path decisions could be blocked correctly but still fail lookup on GET /api/v1/decisions/{id}/explain. v7.1.1 closes that gap so plugin-generated decision IDs resolve through Decision Explainability the same way workflow decisions do.

Active overrides now apply consistently​

The patch fixes override behavior across both major governance surfaces:

  • MCP input/output checks
  • workflow step-gate paths

That matters because a user-visible "override available" hint is only useful if the platform actually honors the override on the next governed request. v7.1.1 makes that true across the full release surface.

Audit search filter parity is real now​

v7.1.0 added decision-, policy-, and override-scoped audit filters. v7.1.1 hardens the backend so audit search no longer drops rows when nullable columns show up in matching events. That keeps Decision Explainability, Session Overrides, and audit reconstruction aligned.

Cache invalidation and policy identity matching were tightened​

The patch also improves two subtle but important behaviors:

  • cached denied workflow decisions are correctly invalidated when a relevant override is created
  • policy identity resolution accepts the forms users actually see in plugin/UI flows, not just one canonical internal representation

For teams using Workflow Checkpoints, this is what makes retry and resume behavior line up with real override-driven recovery.

Per-user workflow scoping now works from X-User-Email​

v7.1.1 also tightens workflow identity handling for plugin-driven runs. Step-gate handlers now fall back to X-User-Email when X-User-ID is absent, which lines WCP scoping up with the identity that plugin clients already forward.

That matters for two practical reasons:

  • per-user cache invalidation works on the identity users actually have in plugin-driven workflows
  • explainability and override-driven retry flows stay scoped to the same caller instead of collapsing into a synthetic shared identity

If you use plugins to drive governed workflows, read this together with Workflow Checkpoints, Decision Explainability, and Session Overrides.

What changed for plugins in the same 3-day release window​

OpenClaw v1.3.1​

OpenClaw is the only plugin that needed a user-facing follow-up patch in addition to the platform fix. v1.3.1 adds config.userEmail, which forwards per-user identity via X-User-Email. This is required for:

  • createOverride
  • revokeOverride
  • listOverrides
  • correct per-user scoping on explainDecision

If you use OpenClaw's override lifecycle methods, upgrade to v1.3.1 and set userEmail in the plugin config. See OpenClaw Integration.

Claude Code v0.5.1, Cursor v0.5.1, Codex v0.4.1​

These follow-up plugin releases did not add new end-user governance features. They added smoke-E2E coverage so the richer block-context and MCP-tool flows are exercised against a live stack rather than only unit tests.

That is worth calling out because it changes confidence, not capability: the v0.5.0 / v0.4.0 feature surface is the same, but the validation story is stronger.

What teams should do now​

If you already adopted v7.1.0, the practical move is simple:

  1. upgrade the platform to v7.1.1
  2. upgrade OpenClaw to v1.3.1 if you use override lifecycle APIs there
  3. upgrade Claude Code, Cursor, and Codex to their latest patch releases for the tested Plugin Batch 1 line
  4. point your internal docs and user guidance at Decision Explainability, Session Overrides, and Workflow Checkpoints

Relationship to v7.1.0​

If you have not read the main release page yet, start with v7.1.0. That page explains the new platform surface itself. This page is the reliability follow-up that tells you which pieces needed patching once real plugin installs exercised the feature set end to end.

Last updated on