What specific models are affected by this silent tool execution failure?

This issue is particularly prominent with the `kimi-coding/k2p5` model, but the underlying parsing problem could affect other models that don't generate tool calls in the exact expected format.

How can I check if my OpenClaw is up-to-date?

If installed globally via npm, run `npm outdated -g openclaw`. If installed via other methods, consult the OpenClaw installation guide. Generally, pulling the latest from the main branch and reinstalling is recommended.

I updated OpenClaw but still have issues. What else can I do?

Ensure you've cleared any old caches or compiled files that might be lingering. If the problem persists, check the specific details in issue #40457 for targeted workarounds or ensure your environment is clean.

Fix: OpenClaw Tool Execution Fails Silently

TL;DR — Quick Fix

Tool execution fails because the agent incorrectly parses tool calls, especially with the kimi-coding/k2p5 model. Update OpenClaw or apply the specific fix.

Run Diagnostics

Next Step

Fix now, then reduce repeat incidents

If this issue keeps coming back, validate your setup in Doctor first, then harden your config.

Open Doctor Harden Config

Error Signal

Agent claims to invoke tools but no tool calls appear in UINo error messages returned, tools appear available but fail to executeread: Claims to read, no file content returned

What's Happening

You're seeing an issue where your OpenClaw agent claims it's executing tools like read, write, or exec, but these actions never actually happen. The UI shows no tool calls, and you get no error messages. This is a regression and breaks core functionality, particularly with models like kimi-coding/k2p5.

The Fix

This is a known regression, and a fix has been merged. The primary solution involves updating your OpenClaw installation. If you're using the kimi-coding model, the specific fix is outlined in issue #40457. This fix recovers downgraded plain-text/XML tool invocations into structured toolCall blocks and avoids treating markdown code snippets as executable calls.

Make sure you are on the latest OpenClaw version. If you need to apply the fix manually before an update, check the code changes in #40457. The change focuses on correctly identifying and parsing tool invocations, especially when they aren't in the expected structured format.

Why This Occurs

This problem arose due to a regression that affects how the agent parses tool invocations. Specifically, the agent was not correctly converting certain tool call formats (often plain text or XML, especially from models like kimi-coding) into the structured toolCall blocks that OpenClaw expects. Instead of recognizing these as valid commands, the agent would essentially ignore them, leading to the silent failure where it claims to have executed a tool but never actually did.

Prevention

Keep your OpenClaw installation up-to-date. Regularly pull the latest changes and update your dependencies to avoid falling victim to regressions. Monitor the OpenClaw GitHub repository for new issues and releases, especially if you rely on specific models or tool functionalities. Test critical workflows after updates to catch regressions early.

Last Updated: March 2026

Did this guide solve your problem?