OpenClaw Cuts Agent Latency and Package Size Through Core Modularization

OpenClawOpenClaw

· Updated

OpenClaw released a performance report showing that its self-hosted AI agents now run nearly 3x faster after a major architectural sweep. By moving heavy provider dependencies into optional plugins, the platform reduced its package size by 59% and its dependency count by 42%.

OpenClaw, an open-source AI assistant for messaging platforms, released a performance sweep showing that cold agent turns are now 2.9x faster. The update follows OpenClaw's modularized runtime architecture that moved heavy provider dependencies out of the core package and into optional plugins.
Cold agent turn speed
2.9x faster (3.4s)
Warm agent turn speed
2.5x faster (3.0s)
Published tarball size
17.8 MB (59% smaller)
Installed dependencies
371 (42% reduction)
Availability
GitHub, Docker, and npm

This optimization addresses the bloat common in self-hosted AI, where adding integrations often slows down the system. By shrinking the package by 59%, OpenClaw lowers the security audit surface (the total code exposed to potential vulnerabilities). This ensures that adding messaging channels or model providers no longer degrades the core agent loop.

You can deploy version 2026.5.27 now via GitHub or Docker. While this release has a known duplicate dependency issue, it maintains the core performance gains and modular structure. The team expects to further reduce the dependency count to 314 in the next release as they refine the plugin extraction process.

View the full update on openclaw.ai
OpenClaw🦞
OpenClaw🦞
@openclaw
X

OpenClaw’s latest sweep: cold agent turns 2.9x faster, warm turns 2.5x faster, tarball 59% smaller, deps down 42% from the monthly high. Small core, explicit deps, optional power in plugins. The claws are getting sharper 🦞 https://t.co/aRvnZjEjUU

47retweets570likes
View on X

Still wondering? A few quick answers below.

OpenClaw agents are significantly faster following recent architectural optimizations. Cold agent turns, which represent the initial startup and execution, are now 2.9x faster, dropping from 9.8 seconds to 3.4 seconds. Warm turns for subsequent actions are 2.5x faster, now taking only 3.0 seconds to complete compared to the previous 7.5 seconds.

The team achieved a 59% reduction in package size by moving heavy provider dependencies out of the core system and into optional plugins. This modularization removed the dependency paths for services like AWS Bedrock, Slack, and Anthropic Vertex from the main core, cutting the total installed dependency count by 42% from its monthly high.

OpenClaw is an open-source, self-hosted AI assistant that users can run on their own hardware. The latest version, 2026.5.27, is available for deployment through GitHub, Docker, and npm. Because it is self-hosted, it allows users to maintain control over their data while automating tasks across messaging platforms like Telegram and Discord.

A smaller core architecture improves performance on resource-constrained hardware and reduces the security audit surface by having fewer active dependencies. By making dependency ownership explicit and moving optional capabilities into plugins, the system becomes easier to trust and maintain while cutting the installation footprint and reducing the risk of native package surprises.

The 2026.5.27 release has a known issue where a bad package shape causes npm to install a duplicate dependency tree, resulting in a larger fresh install footprint than intended. While this duplicate tree is still visible in the current release, the development team has already removed it on the main branch for the next version.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from OpenClaw →

Share this update