Intentional AI Development

Tools, content, and methodology for developers who use AI with purpose -- not on autopilot.

THE BREWCODE FAMILY

Ten products. One philosophy.

Tools, content, and infra - each one usable on its own, all of them better together.

Yasna Production

Voice-first AI assistant with memory — in Telegram

Voice-first AI assistant that remembers your context, reminds you, and acts on your behalf — by voice or text, right inside Telegram.

100+ groups created · 200 users

Open Yasna

BrewPage Production

Publish anything with a single command

Multi-page sites via ZIP or file uploads, single HTML pages, key-value stores, JSON endpoints, and static file hosting. Publish with one request and get a public HTTPS link back in seconds, with no signup and no deploy step.

Open BrewPage API Docs OpenAPI JSON

Brewcode Plugin Production

Plugin suite for Claude Code

A plugin suite that turns Claude Code into a methodical development partner. Split work into bounded units, track it on a file-based task board, and re-anchor the plan after a compaction. Four plugins, 27 skills, 9 agents, 3 hooks -- every skill runs only when you call it. Two mods add typed prompt placeholders and plan-mode effort.

View Docs Mods GitHub

BrewCast AI Active

AI development news, no noise

A curated multi-platform feed covering Claude Code releases, practical tips, and AI development patterns. Telegram (EN + RU mirror), X, and LinkedIn. Short-form, developer-first analysis -- pick the platform you already read.

Join Channel

Brew AI Hub Active Free tier

Self-hosted open-source AI stack — one API for all of it

Self-hosted open-source AI stack — all accessible via one API.

Whisper large-v3-turbo — STT, full transcription pipeline
Qwen 3.5 9B (AWQ) / Qwen 2.5 7B — LLM (text + VLM)
Flux 2 Dev, Flux 2 Klein 9B — image generation
Omni Voice — TTS + voice cloning (UI & API)
OpenAI-compatible /v1/* API, X-API-Key auth

Free tier with window-based rate limits — login, grab a token, ship.

Open Hub Get free token →

Brewcode Blog Active

AI development insights and tutorials

Releases, practical tips, and step-by-step tutorials on Claude Code, plugin development, and AI-assisted workflows. Written for developers who build with intent.

Read Blog

AI Solutions Developing

Custom AI pipelines, integrations, voice-to-code

Custom AI pipelines, integrations, and voice-to-code systems tailored to your stack. Enterprise-ready AI solutions, from architecture to production deployment.

Let's talk

Telegram Bots Active

Utility bots for day-to-day work

Compact Telegram bots that solve one thing well. First up: @zapisposhla_bot — transcribes audio and video to text, subtitles, and brief summary. More bots coming.

See bots @zapisposhla_bot →

Brew Voice Beta

Vibe-code in your native tongue

Voice-to-prompt for vibe-coding with built-in translation. Dictate tasks in RU or EN (or translate between), auto-insert method names and stack terms from managed project context, then ship the prompt to Claude Code / Codex / Cursor / any LLM. Fixed pipeline: 200s of speech → ~6s processed.

See demo Early access Enterprise

brewdoc CLI Production

Document-to-Markdown conversion for agent systems

Standalone command-line tool: fast, deterministic conversion of PDF, DOCX and XLSX files to Markdown. Also converts a public web page on request. No model weights, OCR or external services. A separate tool from the brewdoc plugin of the Brewcode plugin suite.

Open on PyPI GitHub

Claude Code mods

Small mods for Claude Code.

Type ++r, get a whole prompt. Plan harder with /plan-effort.

Explore mods GitHub33+

type this, get that

++rwhole review brief
/plan-effort highplan-mode effort

Philosophy

The future is not vibe coding. It is mindful development with AI agents, where quality matters more than lines of code.

Maksim K.

Task Board

Keep a file-based board with a methodology and base work units for every task. Each active top-level task gets an hourly anti-drift timer while the session runs; change the cadence or opt out.

Bounded Delegation

Every subagent gets one bounded unit and a six-field brief: goal, role, scope, context, consumer, done. After a compaction, a hook re-anchors the plan and the task graph.

Code Review

Generate a deep-review skill tailored to your project: mechanical gates and an intent check by default, domain-expert routing, scope discipline, and adversarial validation at extended depth.

Token Optimization

Reduce token usage in prompts, docs, and rules with depth modes from light to max. Smaller context means faster responses and lower costs.

Secrets Scanning

Scan git-tracked files for leaked credentials, API keys, and tokens using parallel agents. Fixes are offered after the scan and need your approval.

Rules

Check the status of your .claude/rules/ files, list them, and create, improve, or review them. Edits go through a specialist agent.