Fetch.ai has released create-fetch-agent, an open-source command-line tool designed to generate runnable autonomous-agent projects from a single setup command. The release targets one of the most persistent barriers in agent development: configuring communication, identities, ports and dependencies before application logic can be tested. Developers can launch the interactive scaffolder with npx create-fetch-agent, choose a project structure and receive a working Python environment with customization concentrated in designated workflow functions.
The tool currently supports four project structures: a single agent, an ASI chat agent, a multi-agent configuration and a payment-enabled agent that charges in on-chain FET. Each template is delivered as an executable starting point rather than a static code sample, with pre-generated unique seeds, deterministic ports, a Makefile and a dependency manifest matched to the selected Python package manager.
Stop scaffolding AI agents from scratch.
One command gives you four runnable architectures:
→ Single agent
→ ASI:One chat agent
→ Multi-agent systemEvery `create-fetch-agent` project includes:
✓ Pre-generated seeds
✓ Deterministic ports
✓ Agent Chat Protocol
✓…— Fetch.ai (@Fetch_ai) August 4, 2026
Standardized Setup Reduces Manual Agent Wiring
Every generated project includes Fetch.ai’s Agent Chat Protocol and publishes its protocol manifest when it starts. That allows compatible services to recognize how the agent accepts and returns messages without requiring developers to assemble the communication layer manually. The common protocol is intended to make generated agents discoverable and conversational within ASI, while preserving the developer’s control over deployment.
The multi-agent template creates several independent expert agents instead of placing them behind a single local orchestrator. ASI can then discover and route requests among those agents based on their descriptions. This architecture shifts coordination toward the network’s discovery and routing layer, although developers must still define each agent’s responsibilities, behavior and underlying business logic.
The scaffolder also distinguishes project generation from production deployment. It can prepare dependencies, environment files and connection instructions, but it does not automatically publish an agent to Agentverse. That separation limits the promise of “one-command” creation to local scaffolding and execution, rather than implying that a finished autonomous service can be deployed without further configuration, testing or operational controls.
ASI Provides a Separate Creation and Orchestration Layer
Fetch.ai’s Agent Workbench complements the new scaffolding tool but is not the same product. Introduced as an interface inside ASI, the workbench lets creators define an agent’s persona, connect knowledge sources, select a model and test responses in the environment where the agent may later be discovered. The workbench focuses on agent identity and behavior, while create-fetch-agent focuses on generating runnable code and infrastructure.
ASI’s model documentation includes lightweight options for latency-sensitive workloads and extended or agentic variants for deeper reasoning and multi-stage coordination. Its agentic models are designed to search Agentverse, select specialized agents and construct delegation plans for complex workflows. Those capabilities provide the orchestration layer behind the multi-agent templates, but performance still depends on the quality, security and availability of the individual agents being called.
The update gives developers a faster route to experimenting with connected agents, payment flows and ASI services. It does not, however, resolve the harder questions surrounding reliability, permissions and commercial demand. The practical test will be whether standardized scaffolding produces durable applications, rather than simply increasing the number of prototypes that can be launched with less initial setup.