Thomas Sprayberry

dario. A local OpenAI- and Anthropic-compatible proxy that brings your Claude Pro/Max subscription to Cursor, Cline, Aider, and the Agent SDK. Session-affinity routing, multi-account pooling, and live drift detection keep long agent runs stable across tools.

browser-bridge. Stealth headless Chromium in a container. Exposes Chrome DevTools Protocol on 9222. Connect from Playwright, Puppeteer, MCP browser tools, or any agent that wants a remote browser without bundling one.

hybrid. Local-first LLM router: answer the easy majority on a small local model, escalate only the hard queries to any OpenAI-compatible frontier endpoint. ~160 lines, stdlib-only.

askalf. Own Your Stack — open-source agent security (redstamp · truecopy · strongroom) and AI infrastructure you own instead of rent by the token.

deepdive. Local research agent — one command, a cited answer with its trust signals: source-authority mix, lexical citation verification, confidence, cost. CLI + MCP server. Runs on your own LLM router; nightly public report card. Zero hosted dependencies.

hands. Cross-platform computer-use agent. Your LLM on your mouse, keyboard, and screen. Windows (PowerShell), macOS (open + osascript), Linux (xdotool / ydotool). Voice optional, safety guardrails. Routes through dario or any Anthropic-compat endpoint.

amnesia. Search the web. Remember nothing. 155 engines, zero tracking, VPN-tunneled. Self-hosted privacy search via SearXNG. Live at amnesia.tax

agent-security-stack. The open-source agent-security stack — redstamp + truecopy + strongroom — composed into one layered defense. Vet the tool, contain the call, give it a key it never holds. Part of Own Your Stack.

redstamp. A deterministic, offline firewall for AI-agent tool calls — green/yellow/red/black risk tiers, secret-exfil & prompt-injection blocking, tamper-evident audit — plus arena, an open agent-firewall benchmark. Runs as a Claude Code hook or MCP proxy; one-line install from sigstore-signed release tarballs. Part of Own Your Agent Security.

.github. Deprecated mirror — the live profile is github.com/askalf (askalf/askalf). Archived.

install-kit. A curl-pipe-bash template for self-hosted Docker apps. Banner, prerequisite probes, .env scaffolding with crypto-rand secrets, healthcheck wait loop, browser auto-open. Fork it, edit a handful of CONFIGURE blocks, and ship a one-line installer for your own project.

redisflex. One Redis API. Two modes. ioredis for production, in-process Map+EventEmitter for standalone / dev. Also ships a BullMQ-shaped in-memory queue so you can drop the Redis dep entirely when you don't need it.

git-providers. One GitProvider interface for GitHub + GitLab + Bitbucket Cloud (user info, list repos, list branches, OAuth config). Plus a 44-entry api-key-provider taxonomy for cloud / CI / PM / monitoring / e-commerce / marketing / social / productivity / analytics / finance. Zero runtime deps.

pgflex. One Postgres API. Two modes. Real PostgreSQL (pg) for production, PGlite (in-process WASM) for standalone / dev. Same SQL, same query shape, drop the server when you don't need it.

claude-bridge. Bridge Claude Code sessions to Discord. Stay in sync on the go.

brio. The capability layer for AI workloads — semantic cache, cost-aware tiering, structured cost reporting, policy enforcement. Sits in front of any Anthropic-compat endpoint (dario, api.anthropic.com, OpenRouter, vLLM, Ollama). MIT, runs on your laptop.

arnie. Portable IT troubleshooting companion powered by Claude — networking, AD, Windows Update, package managers, log triage, hardware checks, dev environments. Routes through dario for subscription billing.

claude-sync. Sync Claude Code sessions across machines. Pack a local CC session into a portable .ccsync file, ship it via Dropbox / iCloud / Syncthing / USB stick, unpack on the other side. Path-hash mismatches solved via git-remote-url as canonical project key. Zero runtime deps.

whisper.cpp. Port of OpenAI's Whisper model in C/C++

checkout-with-retry. Composite GitHub Action: wraps actions/checkout with auto-retry on transient credential failures (3 attempts, exponential backoff). Used across askalf/* repos to eliminate ephemeral-runner 401 false positives.

mux. Sealed-sender overflow protocol for Claude capacity sharing — trust-group lending with cryptographic unlinkability (RSA blind signatures).

agent. Connect any device to the askalf fleet. WebSocket bridge that dispatches autonomous agent tasks to your machine.

casey. AI IT service-desk CSR — Tier-1 intake + triage. Resolves Tier-1, routes Tier-3 to arnie. Pairs with arnie + dario.

strongroom. Own your agent secrets — your AI agent holds a scoped, single-use lease, never the raw API key. strongroom injects the real secret only at egress, so a leaked prompt or poisoned tool can't leak a credential. Encrypted vault · tamper-evident audit · zero deps · MCP server included. Part of Own Your Agent Security (redstamp · truecopy · strongroom).

fieldpass. own your agent browser — an indirect-prompt-injection firewall + action gate for any CDP browser

cordon. PII-redacting LLM compliance gateway — own your prompts; PII never leaves your perimeter

truecopy. own your agent skills — vet, sign & pin every skill & MCP server before it runs. 68,560 skills scanned, 0 confirmed malicious; weekly watch over the official Claude Code plugin directory.

truecopy-action. Gate every agent skill & MCP server change in CI — the GitHub Action for truecopy, the supply-chain gate for AI agents. Verify truecopy.lock for drift & poisoning, or poison-scan a manifest / marketplace. Part of Own Your Stack.

bring-your-own-model. Run Claude Code on any OpenAI or OpenAI-compatible model — a tiny, dependency-free local bridge that translates the Anthropic Messages API to and from OpenAI (Responses + Chat Completions).