Wayland Core is in final pre-release. Available within 24 hours.
CORE / terminal-first autonomy

Forged to run.Hardened to last.Built to evolve.

A terminal-first agent that plans, runs real tools, asks when it matters, and scales to a swarm — for any kind of work, not just code.

>npx @ferroxlabs/wayland-coreSoon
Activity
type / for commands
Esc interrupt   ⏎ queue a message for the next turn
claude-sonnet-4-6 Default ctx 0% $0.00 00:11
Provider-neutral
MCP client + server
OS-native sandbox
Worktree swarms
Failover + retries
JSON-stream embedding
Self-evolving
Egress firewall
Engine & app

Wayland Core is the engine.
Wayland Desktop is the app.

Core is the open Rust runtime that does the work — the agent loop, tools, providers, sandbox, and memory. Desktop is the GUI product that embeds Core over a JSON-stream protocol. Same engine: run it raw in your terminal, or get the full command center on top.

CORE · you are hereApache-2.0

Wayland Core

The embeddable engine. One Rust binary — terminal-first agent, headless JSON-stream, MCP, OS-native sandbox, ~20 providers, ~75 tools.

  • Run in your terminal or headless
  • Embed it behind your own app
  • Expose it as an MCP server
$ npx @ferroxlabs/wayland-core
DESKTOP · the appGUI product

Wayland Desktop

The polished command center built on Core: assistants, workflows, teams, a sub-agent monitor, approvals, and a UI — for everyone, not just the terminal.

  • Visual workspace & approvals
  • Assistants, workflows, teams
  • Local or cloud, your call
Get the desktop app →
What it does

Three things almost no other agent does together.

Most agent CLIs lock you to one vendor, run tools wide-open, and never get better. Core is built the other way.

Provider-neutral by design

~20 providers — Anthropic, OpenAI, Bedrock, Vertex, Gemini, plus ~15 OpenAI-compatible vendors and local Ollama — behind one trait, where every quirk is config data, not a code branch. Per-request retries, key rotation, a cross-provider fallback chain, and a circuit breaker on every provider. Switching vendors is a one-line change.

~20 providersFailover + retriesHundreds of models via routers

Hardened to run untrusted work

Every shell command runs in the platform's strongest sandbox — bubblewrap, sandbox-exec, AppContainer — and fails closed when none is available. Every outbound request passes a default-on egress gate, enforced structurally by a Clippy lint. Approval modes gate the risky calls; budget caps actually block. Real security, not a checkbox.

OS-native sandboxEgress firewallApproval gatesBudget caps

It learns as it works

An offline evolutionary optimizer (GEPA) mutates and scores skill prompts across generations, and an in-session drafter auto-writes a reusable skill after three successful runs on the same task. Winners seed your library with full lineage — it gets sharper the more it runs. (Offline plus drafting, reviewed before promotion — not a live production self-rewrite.)

GEPA optimizerAuto-drafted skillsCognitive memory
The full runtime

And everything else, built in.

One binary. Point it at a model and it does the work — calling real tools, orchestrating sub-agents, and staying inside your budget.

One binary, three modes

An interactive ratatui TUI, a one-shot wayland-core "prompt", or a headless --json-stream engine — from a single entry point.

~75 built-in tools

Files, shell, search, web, media, scheduling — availability-gated, with a no-stubs contract: env-gated tools hide when credentials are absent.

MCP client + server

Connect any MCP server over stdio, SSE, or streamable-HTTP — or run wayland-core mcp-serve to expose your tools to other hosts.

Sub-agents & swarms

Fan work to parallel sub-agents with Spawn and Delegate, or run worktree-isolated swarms — up to a 100-agent Fleet, each on its own git branch.

ForgeFlows

Declarative, validated workflows — Agent, Pipeline, and Parallel stages — run as a DAG, callable as a tool or straight from the CLI.

Cognitive memory

A real 5-partition × 3-tier memory, SQLite-backed and searchable by meaning. On by default, with background decay. What you teach it survives.

Scheduled tasks

Cron expressions that fire a slash command, a channel message, or a skill — from the CLI, or scheduled by the agent itself.

10 messaging channels

Drive the agent from Slack, Discord, Telegram, Signal, WhatsApp, SMS, Email, Matrix, Teams, or iMessage — one config file.

Plan mode

A read-only mode that studies the code and writes a structured plan before it touches a file. Approve, keep, or discard it.

Plus context compaction, named profiles with inheritance, project-context injection, computer-use on four OSes, a browser tool, an eval harness, and signed self-update.

For builders

Embed the whole agent loop behind your app.

One flag — --json-stream — turns Core into a headless engine you drive over a typed, versioned JSON-Lines protocol. Stream tokens, render tool calls, gate approvals, resume sessions. It's exactly how Wayland Desktop is built.

  • Typed events — text deltas, tool requests & results, sub-agent events, approvals.
  • Honest errors — a retryable flag on every failure, no guessing.
  • Host-driven approvals — pause on a tool; approve once, always, or deny.
  • Apache-2.0 — build and ship commercial software on it.
Read the embedding docs →
$ wayland-core --json-stream "fix the failing test"
{"type":"ready","session":"a1b2c3"}
{"type":"text_delta","text":"I'll find the test…"}
{"type":"tool_request","name":"Bash","input":{"cmd":"cargo test"}}
{"type":"tool_result","name":"Bash","ok":true}
{"type":"text_delta","text":"Fixed an off-by-one."}
{"type":"stream_end","usage":{"in":4200,"out":890}}
Coming soon

Final pre-release. Within 24 hours.
macOS · Linux · Windows.

$npm i -g @ferroxlabs/wayland-core

Wayland Core is in final pre-release hardening. The install goes live within 24 hours. Read the docs now to get a head start.