Block rogue AI agents in application memory before they burn your budget. In-process circuit breaker. No proxy. No network roundtrip. Patent pending.

AeneasSoft is an open-source AI agent observability and active defense tool. It intercepts every LLM API call your agents make — and blocks the dangerous ones before they leave your process.

Two lines of code. Zero configuration. No proxy.

import agentwatch
agentwatch.init()  # Every LLM call is now monitored. Rogue agents blocked in RAM.

pip install aeneas-agentwatch

import agentwatch
agentwatch.init()

# That's it. Make any LLM call:
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello"}]
)
# Trace automatically captured. Cost tracked. Agent defended.

docker compose -f docker-compose.local.yml up -d
# Open http://localhost:3001/health

No account needed. No API key. No cloud dependency.

The killer feature. Block runaway agents in application memory before the request leaves your process:

agentwatch.init(
    budget_per_hour=10.0,        # Alert if agent spends > $10/hour
    max_error_rate=0.5,           # Alert if > 50% errors
    max_calls_per_minute=100,     # Detect infinite loops
    block_on_threshold=True       # Actually BLOCK the request (opt-in)
)

# Per-agent budgets:
with agentwatch.agent("ExpensiveBot", budget_per_hour=5.0, block_on_threshold=True):
    result = client.chat.completions.create(...)
# → CircuitBreakerException if budget exceeded. Request never sent.

How it works: We patch HTTP transport libraries (httpx, requests, aiohttp) at the lowest level. Before every outgoing request, we check budget/error/loop thresholds. If exceeded, we raise CircuitBreakerException — the request never leaves RAM. No proxy. No network roundtrip. No single point of failure.

Your Code
    |
    v
[agentwatch.init()]
    |
    +--→ Layer 1: SDK Patcher (OpenAI, Anthropic)
    |       set_sdk_active(True) → prevents duplicate spans
    |
    +--→ Layer 2: HTTP Interceptor (httpx, requests, aiohttp)
            is_sdk_active()? → skip logging (dedup)
            Active Defense → budget/error/loop check
            |
            v
        AI Provider (OpenAI, Anthropic, Gemini, etc.)

Layer 1 wraps SDK methods for rich structured data. Layer 2 catches everything at HTTP level for framework-agnostic coverage. The deduplication flag (thread-local + ContextVar) ensures no duplicate spans.

Works automatically with: OpenAI, Anthropic, Gemini, Mistral, Groq, Cohere, Together AI, Fireworks, Azure OpenAI, Ollama — and any provider accessible via HTTP.

npm install @aeneassoft/sdk-node

import { init } from '@aeneassoft/sdk-node';
init({ apiKey: 'local' });

docker compose -f docker-compose.local.yml up -d

Starts ClickHouse + Backend. No Kafka. No auth. Dashboard at localhost:3001.

docker compose up -d

Starts ClickHouse + Kafka + Backend + Proxy. Configure via .env.

We believe transparency builds more trust than hiding behind "Beta" labels.

• Streaming: Non-streaming calls captured with 100% accuracy. Streaming calls capture request + final usage summary (tokens + cost), not individual chunks. Full chunk-level tracing ships Q3 2026.
• Cost calculation: Uses list prices for 20+ models. Batch API, cached tokens, and fine-tuned model rates are not reflected.
• Single-process: Each process needs its own init() call. No central proxy config. This is a trade-off of in-process architecture.
• Monkey-patching: We modify HTTP library internals at runtime. When libraries release major versions, our patchers may need updates. We ship SDK updates within 48 hours of breaking changes.

Automated compliance scoring + RSA-2048 signed PDF reports. Available as Enterprise feature. Contact us for details.

MIT License — the SDK, interceptor, circuit breaker, and dashboard are all open source.

The method (Dual-Layer Telemetry Interception and Active Defense) is protected by a USPTO Provisional Patent (April 2026).

Join the Aeneas Community on Discord — get help, share your agent setups, and shape the product:

• Discord: discord.gg/3QjFDQmCJ
• Website: aeneassoft.com
• PyPI: aeneas-agentwatch
• npm: @aeneassoft/sdk-node
• Documentation: aeneassoft.com/docs

"Daa, I love the project" 🦄🛸