ToolExperimentalMCPA2A

wayland

Name: wayland
Rating: 3.1 (609 reviews)
Author: FerroxLabs

by FerroxLabs

Local-first agent runtime that perceives, reasons, and acts with protocol-aware tracing

TypeScript

Updated Sep 19, 2026

609

Stars

113

Forks

View on GitHub

Summary

Implements an agent runtime that perceives, reasons, and acts from a desktop/local-first environment. Uses modular sensors, planners, and executors so agents can gather context, form plans, and execute tasks across local and remote LLMs. Distinctive features include Electron-based UI for local deployments and explicit support for agent communication protocols (Model Context Protocol (MCP)) and multimodal inputs. Also leverages Hierarchical Multi-Agent Pattern to structure multi-agent orchestration.

Key Benefits

As agents take on more complex, delegated workflows, being able to observe what an agent perceived and why it acted is essential for trust. Wayland surfaces perception and reasoning traces from local-first agents, making it easier to inspect decision paths and reproduce behaviors. That visibility helps teams evaluate agent reliability, diagnose failure modes, and build track records for agent-to-agent interactions. This aligns with the Reflection Pattern to surface introspection and justification.

Ideal For

Developers building desktop or local-first multi-agent prototypes who need inspectable perception and reasoning traces for evaluation and debugging. For structured reasoning traces, consider applying the Tree of Thoughts Pattern to organize deliberations.

Use Cases

Inspecting and replaying what an agent perceived to debug wrong actions
Prototyping multi-agent workflows locally with MCP/ACP message flows
Collecting agent traces to build an agent track record or dataset for evaluation

Avoid common pitfalls

Learn what failures to watch for