104 developers joined forces to rewrite the underlying code, and OpenClaw, the world's most popular open-source AI assistant, has been equipped with a "task brain". Even QQ robots can be controlled now. [Guide] 104 developers have collaborated on a major update to OpenClaw, the world's most popular open-source AI assistant, for the first time equipping the AI agent with a task control panel at the "operating system" level, enabling the AI to manage itself, prioritize tasks, and decline requests: the second half of the AI agent competition has begun.
104 people rewrote the underlying code, OpenClaw was equipped with a "task brain", and even QQ robots can be controlled. A month ago, Alexander Feick, an expert from the network security company eSentire, said something on The New Stack that has been worrying the AI Agent community for a long time:
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots The most fundamental flaw in OpenClaw is not a certain checkbox, but rather the lack of a control plane that can express fine-grained trust boundaries.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be managed https://thenewstack.io/openclaw-github-stars-security/
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be controlled Feick issued such a warning only after seeing OpenClaw's adoption had far exceeded the normal level.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even QQ robots can be controlled. His underlying message is clear: the more capable your AI assistant is, the more dangerous it is, because no one can control it.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. At that time, OpenClaw had just set a record: starting from scratch, it took only about 60 days to surpass React's ten-year accumulation with 250k stars, becoming the software project with the most stars on GitHub.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. However, challenges also arose: an AI Agent platform without a scheduling kernel or permission control, the faster it runs, the harder it crashes.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots Just now, the issue of the missing "control panel" mentioned by Feick has been resolved.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots OpenClaw project founder Peter Steinberger (@steipete) officially announced on X the release of OpenClaw v2026.3.31-beta.1.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots https://github.com/openclaw/openclaw/releases
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. This release marks a shift in OpenClaw's background task scheduling, from fragmented accounting to a unified control interface.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", even QQ robots can be controlled. The ambition behind SQLite, the background task control interface is taking shape.
104 people rewrote the underlying code, and OpenClaw is equipped with a "Task Brain", which can even manage QQ robots. How did OpenClaw's "background tasks" work before this update?
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. In fact, it had hardly ever truly "worked".
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. Previously, OpenClaw's background tasks were only accounting tools at the ACP (Agent Client Protocol) level.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. Subtasks have been completed, but the results may not be able to find the parent session; cron timed tasks and CLI background execution go their separate ways; once a task crashes midway, the recovery mechanism is virtually useless.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots This time, version 2026.3.31-beta.1 has done something big: it has unified the four executors, ACP, subagent, cron, and background CLI, into a single SQLite-backed task ledger.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. This implies several aspects of upgrades:
104 people rewrote the underlying framework, and OpenClaw is equipped with a "Task Brain". Even QQ robots can be managed. Firstly, all background tasks now have unified lifecycle management. Heartbeat monitoring, automatic recovery of lost tasks, auditing, and maintenance are all built-in.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain". Even QQ robots can be managed. Secondly, a task flow registry has been introduced. For the first time, developers can use OpenClaw flows list|show|cancel to view and control task flows. Multi-task orchestration has the concept of a "parent record", and is no longer a collection of scattered "orphan processes".
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots Thirdly, blocked tasks can be persisted in a blocked state and retried cleanly on the same flow, rather than being fragmented into new tasks.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain". Even QQ robots can be managed. The results of subtasks can also be traced back to the parent session: after your Agent runs a complex task in the background, it knows which conversation thread to report to.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even control QQ robots. More crucially, this is not just a functional enhancement.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots Kubernetes unified container scheduling into a single control plane, and what OpenClaw has done this time is similar: unifying four different backend execution paths into a single SQLite ledger.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. The difference lies in that Kubernetes schedules containers, while OpenClaw schedules tasks for AI Agents.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even control QQ robots. The founder warned that it's best to upgrade as soon as possible.
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots. This version has 6 Breaking Changes, 4 of which are directly related to security.
104 people rewrote the underlying code, OpenClaw was equipped with a "task brain", and can even manage QQ robots The most important one: the ACP's hazardous tool approval mechanism was rewritten.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even QQ robots can be managed. The previous logic was to override by tool name: adding the tool name to the whitelist would automatically grant access.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even QQ robots can be controlled. The problem is obvious: a tool called "read_file" may have the ability to indirectly execute code, but its name does not reveal the true risk.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", even QQ robots can be managed. Now it has been changed to approval based on semantic categories: only narrow-range read-only operations (search, read) can be automatically approved, while tools with indirect execution capabilities and control plane tools require explicit user confirmation.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots This change targets the ACP approval chain, not a complete replacement of OpenClaw's entire approval system, but it has blocked the largest permission loophole that existed previously.
104 people rewrote the underlying code, and OpenClaw was equipped with a "Task Brain", which can even control QQ robots. This is still not all.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots Plugin installation now defaults to fail-closed: if the built-in security scan detects malicious code, the installation will fail directly.
104 people rewrote the underlying code, OpenClaw was equipped with a "Task Brain", and can even control QQ robots. Want to forcibly install it?
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. You have to manually add the deliberately long parameter: dangerously-force-unsafe-install.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots Gateway authentication is being tightened: trusted-proxy no longer accepts hybrid shared token configurations; node commands remain disabled until pairing approval is granted; tasks triggered by nodes are restricted to a reduced trusted surface.
104 people rewrote the underlying code, OpenClaw was equipped with a "Task Brain", and can even control QQ robots, with these detailed repairs:
104 people rewrote the underlying code, OpenClaw installed the "Task Brain", even QQ robots can be managed Preventing Docker endpoints, TLS trust roots, Python package indexes, and compiler include paths from being overridden by request-level environment variables; blocking approval bypasses for caffeinate and sandbox-exec; detecting inline eval in command carriers such as awk, find, xargs, and make...
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots Contributions from AntAISecurityLab are throughout the Changelog, including path resolution race conditions, early rejection of image bombs, cross-domain redirect cookie leaks, and sandbox symbolic link escapes - each one is an attack surface dug up in actual combat.
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots Steinberger emphasized repeatedly when releasing the beta version: This update mainly focuses on security enhancements, so you really should upgrade as soon as possible.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be controlled Multimodal, multi-terminal explosion, the tentacles of globalized Agents are extending
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even allowing it to manage QQ robots. In addition to the aforementioned framework and nervous system, there are also some "muscle"-level modifications, mainly referring to updates in channel coverage.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be controlled They point to a clear signal: OpenClaw's Agent tentacles are extending globally.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. One of the changes is that QQ Bot has been officially added as a bundled channel plugin.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. It supports multiple account configurations, SecretRef credential management, slash commands, reminder functions, and media sending and receiving.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots The update to WhatsApp seems minor, but the design intention is worth pondering: Agents can now use emojis to respond to messages with expressions: using ❤️ instead of typing out a response.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", even QQ robots can be controlled This is to make chatting with robots closer to "natural conversation".
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots Matrix has added streaming responses: some replies will now update the same message in place, rather than sending a new message for each chunk.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even control QQ robots. LINE now supports sending pictures, videos, and audio externally.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots. Microsoft Teams has added Graph-backed member information queries.
104 people rewrote the underlying code, OpenClaw was equipped with "Task Brain", and can even manage QQ robots. There are also CJK font package repairs.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. Contextual cutting finally no longer underestimates the length of Japanese and Chinese Extension B characters.
104 people rewrote the underlying code, and OpenClaw is equipped with a "Task Brain", which can even manage QQ robots. The memory search function now includes triple-based word segmentation and short CJK substring fallback.
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots The sandbox browser has installed fonts-noto-cjk: the Chinese, Japanese, and Korean characters in the screenshot are no longer garbled blocks
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even control QQ robots TTS automatically detects CJK-dominated text and switches to Chinese voice;
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots. Markdown rendering fixed the issue of bare links devouring adjacent CJK characters.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots These "minor repairs" have a cumulative effect, showing that OpenClaw is moving from a geek toy for an English developer to a "globalized multilingual infrastructure".
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots What it wants is no longer just GitHub Trending, but has set its sights on the broader global market.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. Behind 343k Star, unbridled growth was put on an institutionalized track.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots In addition to technical details, there is another interesting figure: 104 contributors participated in this beta version.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be managed Steipete had previously lamented on Twitter about OpenClaw's "happy troubles":
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", which can even manage QQ robots. PRs are growing at an impossible speed: one day, approximately 600 commits were submitted, so an AI is needed to scan each PR and Issue and remove duplicates.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots Netizen jonahships_ said: "Claw can continuously build new capabilities on top of itself, and all you need to do is talk to it on Discord. The future has arrived."
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even control QQ robots. Netizen rovensky's judgment is more sharp, believing that OpenClaw is the real thing that may kill off a large number of SaaS companies.
104 people rewrote the underlying architecture. OpenClaw now has a "task brain" — it can even manage QQ bots. This is what will actually wipe out a large swath of SaaS startups: not ChatGPT. Because it can be hacked, and more importantly, it can self-hack. And it can be deployed locally. That will crush traditional SaaS.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. In terms of data: as of March 2026, 172 startup companies have built products based on the OpenClaw ecosystem, with monthly ecosystem revenue reaching $361,000.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. ClawHub's skill market grew from 5,700 skills in early February to 13,729. The monthly visit volume reached 27 million, with 2 million monthly active users.
104 people rewrote the underlying code, and OpenClaw was equipped with a "task brain", even capable of managing QQ robots. However, beneath the excitement, security issues have become increasingly severe.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be controlled Over 5,000 open issues. The pressure of reviewing 3,100 commits per day. A Meta executive's Agent accidentally deleted their entire email inbox. A computer science student discovered his Agent had created a profile on a dating website on its own. Security researchers have disclosed zero-click hijacking vulnerabilities...
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots Without a doubt, the open-source community's pace has surpassed the product iteration rhythm of any company. If speed loses order, it will be a disaster.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots. This v2026.3.31-beta.1 version's task control plane and tightened security are essentially using architectural methods to put this uncontrolled growth on a institutionalized track.
104 people rewrote the underlying code, OpenClaw is equipped with a "Task Brain", even QQ robots can be controlled The wave of technological popularization
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", which can even manage QQ robots A month ago, Feick warned that the control plane should be embedded in tools like OpenClaw from the outset, rather than being added later.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", even QQ robots can be managed. The market will push tools like OpenClaw to a position far ahead of governance frameworks: unless we embed the control plane, rather than treating it as an afterthought.
104 people rewrote the underlying code, and OpenClaw is equipped with a "Task Brain", which can even manage QQ robots. This v2026.3.31-beta.1 upgrade is a direct response to this: embedding the first AI task control plane into the AI Agent's governance system.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", even capable of managing QQ robots The background task control interface solves scheduling problems; ACP semantic approval tightens permission control; multi-channel capabilities continue to expand, and these three things combined demonstrate that OpenClaw's governance capabilities have been substantially enhanced in this version.
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", even QQ robots can be managed Looking at the project's history, OpenClaw surpassed React's ten-year Star accumulation in just 60 days.
104 people rewrote the underlying code, and OpenClaw is equipped with a "task brain", even QQ robots can be managed This update means that the OpenClaw project is transitioning from the "explosive popularity" stage to the "infrastructure" stage.
104 people rewrote the underlying code, OpenClaw was equipped with a "task brain", and even QQ robots can be controlled. An open-source community, in just two months, spontaneously built the first task control plane in the AI Agent field, which is also rewriting the industry's rules.
104 people rewrote the underlying code, OpenClaw installed the "Task Brain", and even QQ robots can be managed. In the past, the answer to how a product should be defined belonged to product managers of large companies.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots Now, this issue is being taken over by more global contributors like those on GitHub, an open-source community.
104 people rewrote the underlying code, OpenClaw was equipped with a "task brain", and even QQ robots can be controlled A trend of technological democratization is coming, accompanied by the AI wave, more strongly and irreversibly.
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", even QQ robots can be controlled
104 people rewrote the underlying code, OpenClaw is equipped with a "task brain", and can even manage QQ robots
104 people rewrote the underlying code, OpenClaw is equipped with "Task Brain", and can even manage QQ robots https://github.com/openclaw/openclaw/releases
104 people rewrote the underlying code, OpenClaw equipped with "Task Brain", can even manage QQ robots https://thenewstack.io/openclaw-github-stars-security/
104 people rewrote the underlying code, OpenClaw was equipped with a "task brain", and even QQ robots can be controlled