at_think
Write a thought into a strand. Persistent storage receives caller-supplied ciphertext/nonce fields plus an ed25519 signature; it has no plaintext thought column or decrypt path, but does not prove AES-GCM encryption. Self processing stays user-side; bridged hosted runtimes process plaintext in Ag...
This record as markdown: /tools/dev-agenttool-agenttool/at-think.md
What at_think does on Agenttool
AI agents invoke at_think to trigger actions in Agenttool. 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.
Why at_think is rated High
at_think triggers real processes with real consequences. An agent gone sideways doesn't fire it once. It starts dozens of builds, sends mass notifications, or burns through compute before anyone looks up.
Attacks that exploit this kind of access
The rule that runs at_think safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Agenttool, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For at_think, this is the rule to start with:
at_think stays usable, but rate-capped: a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Agenttool, apply this rule, and every at_think call is checked against it from then on.
Questions about at_think
Write a thought into a strand. Persistent storage receives caller-supplied ciphertext/nonce fields plus an ed25519 signature; it has no plaintext thought column or decrypt path, but does not prove AES-GCM encryption. Self processing stays user-side; bridged hosted runtimes process plaintext in AgentTool worker memory. Trusted is experimental: it requires configured platform KMS, uses platform-wrapped runtime key material, and plaintext can enter AgentTool worker memory and the chosen model provider. Provisioning does not run it; explicit POST /v1/runtimes/:id/start is required before its first invitation, after which trusted cycles can persist signed thoughts. Output is one-line: OK thought seq=<n> · <short-id> on /<strand>. It is categorised as a Execute tool in the Agenttool MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Agenttool MCP server in PolicyLayer and add a rule for at_think: 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 Agenttool. Nothing to install.
at_think 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 at_think 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 at_think. 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.
at_think is provided by the Agenttool MCP server (https://api.agenttool.dev/v1/mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Agenttool, and thousands of servers like it.
Across the catalogue