S.DevLabs — Strategic Development Labs. Building what's next. AI, data, P25, beyond.

Exploring technology through independent development, experimentation, and open-source innovation. Built as a hobby, developed with the assistance of AI, and shared with the community.

  • 5libraries
  • MITlicensed
  • Pythoncore stack
  • 0usage fees

Writing

Articles

Insights from independent research, technical development, and experimentation. Documenting discoveries, lessons learned, practical implementations, and findings worth sharing with the community.

Repositories

Projects

Five Python libraries, all MIT licensed, all usable independently. Install from source; nothing here asks for an account or a key.

On the XRP Ledger

Identity, credentials, and audit on-chain 3 projects

xrpl_agent_id

Identity

Give AI agents a verifiable on-chain identity. W3C DID (did:xrpl) plus XLS-70 Verifiable Credentials on the XRP Ledger — issue, accept, and verify. ERC-8004-style, without the EVM.

  • Python library — no extra infrastructure
  • W3C DID method and XLS-70 credentials
  • Free to use — only ledger transaction fees apply
xrpl_agent_id — issue & verify testnet
pip install "xrpl-agent-id @ git+https://github.com/DuNorth595/xrpl_agent_id.git@v0.5.4"

xrpl_agent_log

Audit

Tamper-evident append-only history log for AI agents on the XRP Ledger. Hash-chained, mutually signed, replayable from public data alone.

  • Audit log and observability for autonomous agents
  • Hash chain plus mutual signing
  • Reconstructible from public ledger data only
xrpl_agent_log — hash chain & anchor tamper-evident
pip install "xrpl_agent_log @ git+https://github.com/DuNorth595/xrpl_agent_log.git@v0.4.0"

xrpl_did_mcp

MCP

MCP server for did:xrpl — W3C-compatible agent identity on the XRP Ledger, exposed to any Model Context Protocol client.

  • Model Context Protocol server
  • W3C DID and XLS-70 credential support
  • Drops into an existing agent toolchain
xrpl_did_mcp — tool call MCP
pip install "xrpl-did-mcp @ git+https://github.com/DuNorth595/xrpl_did_mcp.git@v0.6.0"

Agent Stack

Memory, and a voice you can interrupt 2 projects

ai_agent_chat

Voice

Full-duplex voice assistant you can talk over. Talk while it is speaking and it keeps going, catches what you said, and answers once it finishes. Push-to-talk over your phone. Local Whisper STT and Piper/Kokoro TTS.

  • Runs on your own hardware
  • Reachable over a VPN (Tailscale) — no cloud dependency
  • Self-hosted, open-source, Python/FastAPI
ai_agent_chat — full duplex barge-in
idle
git clone https://github.com/DuNorth595/ai_agent_chat.git

ai_agent_memory

Memory

Long-term memory for AI agents that survives between sessions. One index file, one pointer per topic, everything else plain Markdown.

  • No database required
  • Portable, git-friendly, framework-agnostic
  • Plain Markdown you can read and edit by hand
ai_agent_memory — the four layers long-term memory
git clone https://github.com/DuNorth595/ai_agent_memory.git

Open

Ideas & Submissions

Three threads running at once — verifiable identity on a public ledger, AI that lives on your own devices, and agent capabilities that can be reviewed and shared. Plus an open door for anything else.

Building now

Building on the XRP Ledger

Identity, credentials, and tamper-evident audit on a chain that settles in seconds. The libraries are released; the next layer is making them compose — agents that hold verifiable identity and can prove what they did, without asking anyone to trust them.

xrpl_agent_id · xrpl_agent_log · xrpl_did_mcp

Exploring

Mobile apps and AI integration

Phones that talk to the network on your behalf — push-to-talk agents you can interrupt, radio and dispatch tools that live in your pocket, and assistants that run on-device instead of in someone else's cloud. Everything still yours when the connection drops.

iOS · Android · on-device inference

Exploring

AI agent skills

Skills as portable, reviewable, shareable units — a directory an agent can read and another can trust. The same discipline that made a library worth installing: clear scope, a way to verify it, a way to take it apart.

Skills · MCP · verifiable capabilities

Open

Just some ideas?

A ledger nobody has built yet, a tool that should exist, a problem you keep working around, a dataset worth sharing, or a collaboration that would be better with two people on it. No pitch required — a sentence is enough to start.

Ideas · Projects · Collaboration

Submit an idea Ideas, projects, collaborations, or something you want built — all welcome.

Get in touch

Contact

Bug reports, ideas, questions, or a correction to something I've written — all welcome.

No tracking, no newsletter, no CRM — it lands in a normal inbox.