Make predictions using a neural model Use when nothing native trains on your workflow — Claude Code has no learning loop. Use to train SONA/MoE/EWC patterns from successful task outcomes; query via neural_predict before spawning agents. Off-path for one-shot work.
AI agents invoke neural_predict to trigger actions in Claude Flow. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
The tool both queries a neural model for predictions (Read-like) and trains/updates model patterns from outcomes (Write/Execute). Training a model mutates internal state and can influence downstream agent-spawning decisions, making this more than a pure read. The most severe applicable category is Execute, since it triggers model training operations and influences orchestration behavior.
From the tool's definition 'Make predictions using a neural model' and 'train SONA/MoE/EWC patterns from successful task outcomes; query via neural_predict before spawning agents'
Attacks that exploit this kind of access
Make predictions using a neural model Use when nothing native trains on your workflow — Claude Code has no learning loop. Use to train SONA/MoE/EWC patterns from successful task outcomes; query via neural_predict before spawning agents. Off-path for one-shot work. It is categorised as a Execute tool in the Claude Flow MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Claude Flow MCP server in PolicyLayer and add a rule for neural_predict: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Claude Flow. Nothing to install.
neural_predict is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the neural_predict rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for neural_predict. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
neural_predict is provided by the Claude Flow MCP server (claude-flow). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.