Here's what's in the pipeline for BotGuardian OS. Priorities shift based on what you all need most — drop feedback in the chat.

v1.1 (Next Update)

• Team mode — shared ledger for multi-user setups

• Slack/Discord alerts — get kill-switch notifications in your workspace

• CSV export — dump your ledger to spreadsheet for accounting

v1.2

• Per-agent budgets — set individual spend caps for each agent in a multi-agent system

• Auto-scaling thresholds — dynamic limits that adjust based on historical usage patterns

• Docker image — one-line deploy for server-side agent monitoring

v2.0 (Major)

• Cloud sync (optional) — encrypted ledger backup with cross-device dashboard

• Cost forecasting — ML-powered predictions for your monthly AI spend

• API marketplace integration — monitor non-LLM API costs (Serper, Browserless, etc.)

Request a Feature

Reply to this post with what you need most. Upvote others' suggestions. The most-requested features get built first.

BotGuardian OS v1.0 is here. Everything you need to stop runaway AI agents from draining your wallet.

What's Included

🛑 Emergency Kill-Switch

• Real-time spend monitoring across OpenAI, Claude, Gemini, and 20+ LLMs

• Automatic agent termination when thresholds are breached

• Sub-second response time — kills the process before the next API call lands

📊 Local SQLite Ledger

• Every API call logged with model, tokens, cost, and timestamp

• Zero cloud dependency — your data stays on your machine

• Query your spend history with standard SQL

📈 Observability Dashboard

• Per-model cost breakdown

• Burn rate tracking with trend visualization

• Threshold configuration UI

🔌 Framework Support

• n8n (webhook integration)

• LangChain (Python decorator)

• CrewAI (middleware)

• Raw OpenAI / Anthropic SDK wrappers

• Any HTTP-based LLM API via proxy mode

Installation

unzip BotGuardian_OS_v1.0.zip && cd botguardian-os
pip install -r requirements.txt
botguardian init
botguardian watch --threshold 5.00

Known Limitations

• Windows support is CLI-only (dashboard requires WSL)

• Streaming responses are metered at completion, not per-token

• Max 50 concurrent agent sessions in v1.0

Report bugs in the Agent Builders Chat. Ship safe. 🛡️

Most devs look at token pricing and think they know what their agent costs. They're usually off by 30-60%.

Here's why: retries.

When your agent hits a malformed JSON response, a rate limit, or a context window overflow — it retries. Sometimes 3-4 times before succeeding. Each retry burns the same tokens as the original call.

A simple calculation:

• 500 agent tasks/day

• 2,000 avg input tokens, 800 avg output tokens

• Claude Sonnet at $3/$15 per 1M tokens

• 15% error/retry rate (lower than most production agents)

Expected monthly cost: ~$1,800

Actual monthly cost with retries: ~$2,500+

And that's the normal case. The nightmare scenario is an agent stuck in a retry loop overnight with no ceiling. I've seen solo founders wake up to $4,000+ surprise bills from a single bad run.

The fix isn't complicated:

1. Log every token to a local ledger (not a cloud dashboard you check once a week)

2. Set a hard spend ceiling per agent, per hour

3. Wire an automated kill-switch that revokes API keys when the ceiling hits

That's what I built BotGuardian OS to do. Three zero-dependency Python modules. One local SQLite file. An OS-level kill-switch. $29 one-time, runs entirely on your machine.

But even if you don't use my tool — at minimum, audit your retry rate. It's probably higher than you think.

Welcome aboard. You now have access to the full BotGuardian OS toolkit.

Getting Started

1. Download — Head to the Downloads section and grab the latest ZIP

2. Install — pip install botguardian (or unzip and run locally)

3. Wrap your first agent — Add the @guard(threshold=5.00) decorator to any LLM call

4. Launch the dashboard — botguardian dashboard --open

What's Included

• Interceptor — Python wrapper that logs every token to a local SQLite ledger

• Dashboard — Auto-generated HTML/JS visualizer for per-model, per-agent spend

• Kill-switch — Severs API connections + Discord webhook when spend crosses your threshold

• 20 model presets — OpenAI, Anthropic, Google, xAI, DeepSeek, Meta, Mistral, and more

Support

Drop any questions in the Agent Builders Chat. We're here to help you ship agents with confidence.

All future updates ship here automatically — lifetime access, no subscription.