Documentation

How Maylyn works

The command reference and the core concepts. The deep guides live in the repo and in the app, so they never drift from the code.

CLI

Command reference

Run maylyn help, or maylyn <cmd> --help for a command's own flags.

maylyn help
maylyn launch the interactive TUI (default) maylyn serve HTTP/SSE server + web & phone UI (see the API reference) maylyn hub run a EchoBark coordination hub (identity + directory, no model) # Slash commands (TUI + Web) /apps /browse /chart /echobark /gallery /image /library /macro /plan /project /prompt /remind /roll /search /sound maylyn slash list all built-in slash commands # Accounts & platform maylyn users <cmd> list / ban / unban / rm / prune accounts, access codes maylyn status hung, orphaned, or GPU-squatting processes? maylyn config model-server startup and settings maylyn start | stop [unit] background services (daemon / serve / all) maylyn server-status the managed server and its live binary identity maylyn server-restart restart the managed server with the installed binary maylyn service <cmd> install / uninstall / status the boot-time units maylyn genkeys generate access keys maylyn vet self-check + repair (also runs on every launch) maylyn telemetry per-turn drift report (route mix, latency) maylyn telemetry-benchmark cross-interface benchmark (TUI core vs web SSE) # Evaluation maylyn eval --gold FILE score a gold set (--route-only for offline) maylyn eval gen draft a gold set from stored transcripts maylyn eval synth generate a synthetic categorized corpus # Knowledge bases maylyn kb list list knowledge bases (sources, chunks) maylyn kb search <query> see exactly what a turn would retrieve maylyn kb <cmd> show / new / add / reindex / rm maylyn project status is this directory a known project? which KB does it use? maylyn project <cmd> link / kb — set this directory up as a project maylyn vault <cmd> init / status a vault layout in the sandbox # Maintenance maylyn setup install llama.cpp + a model (--voice, --canvas-art) maylyn lora <cmd> install / inspect / manage local LoRA styles (alias: loras) maylyn connectors manage MCP connectors maylyn schedule <cmd> manage scheduled agents maylyn update self-update maylyn prune remove runtime residue maylyn version print version
Extending Maylyn

Connectors, skills & slash commands

Three ways to add capability. They compose — a skill can tell the model when to reach for a connector's tools.

tools

MCP connectors

Maylyn speaks the Model Context Protocol, so any MCP server works. Tool calls still pass the approval gate and run sandboxed.

maylyn connectors status
know-how

Skills

A SKILL.md of task instructions the model loads only when a task needs it. Follows the Agent Skills spec.

cp -r pdf-forms/ ~/.maylyn/skills/
verbs

Slash commands

Built-in verbs that mean the same thing in the terminal and the web. maylyn slash lists them all.

/search webgpu compute shaders
No API keys

Built-in connectors

websearch
docs → Markdown
wikipedia
calculator
time
papers
hackernews & reddit
maps

Also in the catalog when their host tool is present: playwright, chrome-devtools, database, docker, media (yt-dlp), github, context7, memory, filesystem, sequential-thinking. The Library's MCP tab also offers youtube-transcript, slack, notion, google-maps, email and yahoo-finance; those that need credentials ask for them at install.

Any server

Add an MCP server

~/.maylyn/mcp.json
{ "servers": { "my-tool": { "command": "npx", "args": ["-y", "@acme/mcp-my-tool"], "env": { "MY_TOOL_TOKEN": "…" } } } }

The sidecar only learns tool names and schemas — every call RPCs back into Maylyn's Go registry, so sandbox containment, the safety kernel, and the approval prompt are identical to a built-in tool.

Models

Providers

Local by default. Any OpenAI-compatible server works; add a hosted key any time.

providers
llama.cpp · LM Studiolocal & free
Groq · Geminifree tier
OpenAI · Anthropic · DeepSeekpay as you go
OpenRouter200+ models
Project context

MAYLYN.md

Drop a file in a project directory; Maylyn loads it every session started from there.

my-api/MAYLYN.md
## Project: my-api - Language: Go 1.25+ - Style: gofmt + golangci-lint - Tests: go test ./... + net/http/httptest - Never log with fmt.Println; use slog

Two files stack: this per-project MAYLYN.md (standing rules, read from the workspace root) and a NOTES.md scratchpad the model can append to.

Voice & audio

Talk to it

Local text-to-speech and speech-to-text. Enable it at setup.

terminal
maylyn setup --voice # Piper-first TTS with an edge-tts / system fallback, # plus local Whisper for speech-to-text.