Just 24 hours after OpenClaw 3.7 dropped, version 3.8 is already here and it is turning heads across the developer community. The ai porn video generator team behind this open-source AI agent framework clearly is not sleeping. With the release of ACP open-source protocol, a new backup system, Telegram notification fixes, and over 12 security patches, OpenClaw is proving why developers worldwide are calling it the most exciting project in the AI agent space right now.
The momentum is unstoppable. OpenClaw has entered full sprint mode.
Between late March 8 and early March 9, OpenClaw 3.8 went live.
The speed of these releases is shocking the industry. Some developers joke that the global programmer community is pulling all-nighters just to keep up.
One sentence to sum it up
The core updates in this release are four major items:
1. ACP protocol goes open-source, solving the who is talking problem for good
2. Built-in backup and restore via openclaw backup command
3. Telegram notification spam bug finally fixed
4. Over 12 security fixes, many of which the community never even noticed
Even the official team admits there are too many changes to list in one place. It is a lot.
Highlight one: ACP open-source protocol, the identity layer for AI agents
This is a big deal. Before this update, when multiple AI agents talked to each other, nobody really knew who had what permissions or what level of trust existed. It was like a group chat where everyone is anonymous and you have no idea who is making decisions.
With version 3.8, that changes completely.
OpenClaw has open-sourced the ACP protocol, which stands for Agent Communication Protocol. Now every instruction carries a digital signature and identity verification. You know exactly who sent what, which agent made which call, and whether they had permission to do it.
For regular users, this means three critical improvements:
First, collaboration becomes trustworthy because agents cannot pretend to be someone else.
Second, permission control is much finer-grained, so sensitive information stays locked down.
Third, the entire system becomes auditable. Who did what, when, and why is visible at a glance.
Previously this was a major pain point for enterprise teams. Now it is simply solved. For AI security, this is a game-changer.
Highlight two: openclaw backup, a long-awaited feature
This is something users have been asking for since the early days.
Before 3.8, managing OpenClaw configurations felt risky. One wrong move and your entire workspace could break. Many developers wrote their own backup scripts, but the official team never provided a standard way to do it.
Version 3.8 fixes this gap for good. The new backup command is simple and powerful. It supports –only-config to back up just your settings, and –no-include-workspace to exclude workspace data. These options give you full control over what gets saved.
The backup files include manifest validation, so you can restore everything with a single zip file. Even better, the system detects dangerous operations during restore and warns you before overwriting anything. This means you can experiment freely without fear of breaking your setup.
Highlight three: Telegram notification spam killed
This was one of the most annoying bugs for teams using Telegram alerts.
The problem was simple but frustrating. When the same alert fired repeatedly, the system would spam your phone with endless notifications. It did not matter if you muted it or changed settings. The flood just kept coming.
Version 3.8 completely fixes this bug with smarter logic. Now Telegram channels only send one notification per unique event, and duplicate alerts are automatically suppressed. Your phone stays quiet unless something truly new happens.
Highlight four: 12 plus security patches, fixing hidden risks
The official team did not publish a full list of security details, which is standard practice for responsible disclosure. However, they confirmed that over 12 security issues were patched in this release.
Some of these fixes address vulnerabilities that most users never noticed but could have been exploited. The OpenClaw team takes security seriously and treats it as a core feature, not an optional add-on.
For production teams, this is huge. You can deploy with confidence knowing the framework has been hardened against known threats.
Other improvements worth mentioning
Some users felt OpenClaw was too fast and too intense.
The clothing remover ai team responded with a clever analogy. They said an open-source project moving fast is like a train leaving the station. You can either jump on now or watch it leave. The choice is yours.
The speed is actually a signal. It shows the team is active, the community is engaged, and the project is heading somewhere exciting.
So the lobster is not sleeping.
And neither are we.
CADOAN is a professional, independent AI industry blog and information platform dedicated to the research, sharing, and popularization of artificial intelligence. We are a team of AI enthusiasts, researchers, and technical writers who focus on the development and application of modern artificial intelligence. We do not represent any commercial institution, technology company, or AI model camp. Our only position is to provide real, objective, and valuable AI content for readers, learners, developers, and business practitioners around the world.