Agents That Write Their Own Tools at Runtime — Sandhya Subramani, AWS
- Source
- youtube.com
- Author
- AI Engineer
- Date
Context
An can extend its available actions by writing and loading a new tool during a task. A conference talk from Amazon's AWS describes this with the Strands Agents SDK: in the talk description, an editor creates tool code, a shell can execute commands, and load_tool adds the resulting tool to the running agent. Instructions tell the model what a suitable tool should look like. See the talk description.
The project's tool library documents dynamic loading of custom tools with load_tool. This provides a concrete route from generated code to an action the agent can call. The library now marks its editor and shell tools as deprecated in favor of newer built-in replacements, so the talk's exact tool names may not match current usage. The starting point still includes the capabilities needed to create and load tools; the application-specific tools are what the agent adds as it works.
- In AWS developer advocate Sandhya Subramani's demo with the open-source Strands Agents SDK, an agent that starts with zero tools writes a calculator and a character counter when asked, then uses them without restarting.
- The same approach extends to sub-agents: a travel-planner demo assembles its own sub-agents, and the talk walks through swarm, graph, handoff and workflow patterns.
- Proposed for self-modifying agents check goal success, tool choice, parameters and inter-agent flow, with of execution, constrained permissions and observability.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
- sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
- guardrails — The checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.
Checking sign-in…
Loading comments…






