Remote OpenClaw Blog
ACPX 0.6.0: System Prompt Controls, Session Pruning, and Persistent Session Fixes
4 min read ·
ACPX 0.6.0 is the kind of release that tells you what the maintainers think the real pain points are. The headline items are not flashy demos. They are control surfaces, persistence fixes, cleanup tools, and safer queue behavior. That is a strong signal about the maturity problems ACPX is trying to solve.
The Headline Changes
| Area | What 0.6.0 added |
|---|---|
| Claude control | --system-prompt and --append-system-prompt global flags |
| Session cleanup | sessions prune with dry-run, age filters, and optional history cleanup |
| Embedding/runtime | startTurn(...) handles for observing events and controlling turn lifecycles |
| CLI/ACP | --no-terminal to disable advertised ACP terminal capability |
the ACPX 0.6.0 release notes reads like a release for people already using ACPX seriously. The new surface area is about control and cleanup, not a brand-new beginner mode.
Why the Persistent Session Fixes Matter
The fixes section is where the release becomes more interesting. ACPX 0.6.0 keeps reusable persistent clients warm across turns, reuses saved persistent sessions for runtime controls, persists prompts at turn start so history does not disappear mid-run, and improves session status reporting when queue owners are not running.
Those are all symptoms of the same underlying challenge: once sessions are persistent, you need the runtime to behave consistently under reuse, fallback, and recovery. This release is clearly tightening that part of the stack.
WSL Translation, Queue Hardening, and Error Hints
The release also fixes WSL cwd translation when Windows executables are spawned from WSL, tightens queue and IPC directories to owner-only permissions, uses cryptographically random generation IDs for queue owners, and adds clearer text-mode remediation hints for auth failures, timeouts, rate limits, missing sessions, and invalid model names.
Stable Session Layer
If the real problem is session control, recovery, and reuse, start with the durable session layer instead of patching the same runtime pain over and over.
This is the part that makes the release feel operationally serious. It is not just adding convenience flags. It is reducing the number of silent footguns in real runtime use.
Who Benefits Most from 0.6.0
If you embed ACPX or use it for repeated agent work, this release is much more meaningful than if you only open occasional short sessions. Session pruning, runtime turn handles, queue hardening, and error remediation are all features that matter once a workflow has been running long enough to become messy.
That is why the release reads more like an operator-quality upgrade than a marketing release. It is making the runtime easier to keep clean and harder to misuse.
Primary sources
- the ACPX 0.6.0 release notes, for the complete 0.6.0 changes and fixes list.
- the ACPX repository, to confirm the release sits on an actively maintained public codebase.
- the ACPX README, for baseline context on what ACPX is meant to control.
Recommended products for this use case
- Session Supervisor — Best fit if your immediate pain is keeping agent sessions stable, resumable, and reviewable.
- Persistent Dev Orchestrator — Best fit if your next problem after session stability is background orchestration across multiple tasks.
- Operator Launch Kit — Best fit if you want to package a custom operator path around ACPX instead of improvising the file structure yourself.
Limitations and Tradeoffs
This article is anchored to the 0.6.0 release notes on April 25, 2026. ACPX is shipping actively, so some pain points in this post may improve further or shift as later releases land.
Related Guides
- What Is ACPX?
- How ACPX Controls Codex and Claude via Agents
- OpenClaw vs Codex for Long-Running Agent Workflows
- What Is OpenClaude?
FAQ
What are the biggest user-facing wins in ACPX 0.6.0?
The most meaningful wins are session pruning, stronger persistence behavior, cleaner queue safety, WSL path handling, and clearer error hints.
Why are system prompt controls such a big deal?
Because they let callers replace or append Claude system prompt content while staying inside persistent ACPX sessions instead of falling back to ad hoc session resets.
Is this mainly a developer release?
Yes. It is squarely aimed at people already running ACPX seriously enough to care about session lifecycle, embedding, runtime controls, and queue behavior.