Issue with OpenClaw on AWS: Intermittent "Tool not found" & sessions_spawn permissions
Hi OpenClaw Community,
I am currently running an OpenClaw instance on AWS for tasks involving Obsidian management and sub-agent orchestration. I’m facing two critical issues that are causing significant latency and failure rates.
1. Intermittent "Tool not found" errors I’m seeing frequent but inconsistent Tool not found errors for core tools (read, write, edit, exec, memory_search).
Symptoms: A tool fails in one attempt and works perfectly in the next.
Context: This leads to constant retries, making the agent extremely slow. I've already double-checked HEARTBEAT.md, but the issue persists.
2. sessions_spawn AgentId Restriction When the agent attempts to spawn sub-agents using a specific ID (e.g., sessions_spawn(agentId='sally', ...)), it fails with:
Error: agentId is not allowed for sessions_spawn (allowed: none)
Config Checked: I have verified that tools.agentToAgent.enabled is true and tools.agentToAgent.allow includes "sessions_spawn".
Workaround: Currently, I have to spawn without an ID and prompt the persona manually, which is inefficient.
My Hypothesis: It suspects that a global tools.allow configuration in the root openclaw.json might be overriding agent-specific permissions. Since this is on AWS, I’m also wondering if there are any known network/IAM constraints affecting the daemon's communication.
Has anyone experienced similar instability or config conflicts on cloud deployments? Any advice on how to correctly configure openclaw.json to allow agentId assignment would be appreciated.
Thanks in advance.
I am currently running an OpenClaw instance on AWS for tasks involving Obsidian management and sub-agent orchestration. I’m facing two critical issues that are causing significant latency and failure rates.
1. Intermittent "Tool not found" errors I’m seeing frequent but inconsistent Tool not found errors for core tools (read, write, edit, exec, memory_search).
Symptoms: A tool fails in one attempt and works perfectly in the next.
Context: This leads to constant retries, making the agent extremely slow. I've already double-checked HEARTBEAT.md, but the issue persists.
2. sessions_spawn AgentId Restriction When the agent attempts to spawn sub-agents using a specific ID (e.g., sessions_spawn(agentId='sally', ...)), it fails with:
Error: agentId is not allowed for sessions_spawn (allowed: none)
Config Checked: I have verified that tools.agentToAgent.enabled is true and tools.agentToAgent.allow includes "sessions_spawn".
Workaround: Currently, I have to spawn without an ID and prompt the persona manually, which is inefficient.
My Hypothesis: It suspects that a global tools.allow configuration in the root openclaw.json might be overriding agent-specific permissions. Since this is on AWS, I’m also wondering if there are any known network/IAM constraints affecting the daemon's communication.
Has anyone experienced similar instability or config conflicts on cloud deployments? Any advice on how to correctly configure openclaw.json to allow agentId assignment would be appreciated.
Thanks in advance.
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
85,605Members
Resources
Similar Threads
Was this page helpful?
