Agentic Ledger: Open Source Proxy Tool Tracks AI Agent Costs and Loops

·1 views

A developer has released Agentic Ledger, an open source transparent proxy tool designed to monitor AI agent activity without requiring any code changes or SDKs. The tool sits between an AI agent and its LLM provider, recording every HTTP request and response, calculating real costs, and storing data locally in SQLite or Postgres. It includes loop detection to identify repetitive or stuck agent behavior, and can act as a circuit breaker by returning HTTP 429 errors to prevent runaway spending. Hard budget caps can be enforced per session, per agent, or per day before calls even reach the provider. The project is MIT licensed, available via pip or Docker, and the developer is actively seeking testers and contributors to expand its real-world coverage.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Silent Go bug blocked dead push subscription cleanup, causing endless wasted requests

A developer working on agentrq, a Go backend for an AI agent task manager, discovered a logic bug in the web push notification path that had never triggered any error or alert. The flaw stemmed from a misunderstanding of Go's net/http contract: a 410 Gone response is treated as a successful request, so the error-handling branch meant to delete expired subscriptions was never reached. Because the cleanup code was unreachable from the start, stale push subscription records accumulated in the database without any upper bound. Every subsequent notification event continued attempting delivery to already-dead endpoints, adding a small but permanent overhead per dead subscription per event. The bug caused no crash or visible failure, making it invisible to standard monitoring while silently degrading system efficiency over time.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Writer Tests 7 AI OSINT Tools on Himself, Exposes Full Profile in 4 Minutes

A privacy-conscious writer ran seven AI-powered open-source intelligence tools against his own digital identity in 2026 to audit his personal data exposure. Using only his real name, primary email, and a profile photo, the tools collectively surfaced his current address, date of birth, six email addresses, 12 social media profiles, and passwords from three previously unknown data breaches within four minutes. Facial recognition tools found 47 photos of him online, including one sourced from another user's Venmo account, while breach databases linked his email to an old phone number later used to identify personal contacts. The total cost for a malicious actor to replicate this information gathering ranged from zero to 47 dollars depending on the tools used. The writer concluded that even users who practice standard security measures like two-factor authentication and password managers remain highly vulnerable to aggregated public data exposure.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Zero Trust Security Must Extend Into AI Pipelines, Not Just Networks

Enterprise AI systems face unique security risks because large language models retrieve information based on semantic similarity rather than user permissions, making traditional network-level security insufficient. The newly formed Open Secure AI Alliance has flagged fragmented guidance and inconsistent controls as key challenges when organizations integrate LLMs into enterprise environments. Security architects are urged to enforce Zero Trust principles at every stage of the AI pipeline, including data ingestion, embedding, vector storage, retrieval, and output. Practical controls include role-based access filtering at the vector database layer, classification and sanitization of sensitive data before embedding, and ensuring the context window passed to the model contains only information the requesting user is authorized to view. Without these pipeline-level controls, LLMs can inadvertently surface confidential data or become entry points for prompt injection attacks.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Rebuilds 1990s Panasonic Video Mixer as a Browser App Using WebGPU

A software developer has recreated the Panasonic WJ-MX50, a desktop digital A/V mixer from the early 1990s, as a browser-based application called web-mx-50. The original hardware was designed for wedding videographers and cable-access studios, but the author used it extensively as a live VJ instrument in techno clubs through the 2000s. The browser recreation was built using vanilla TypeScript, WebGPU for video processing, and Web Audio, totalling around 5,700 lines of code with no external frameworks. The project used the original 40-page Panasonic operating manual as a formal specification, with the developer's own muscle memory acting as an acceptance test for behavioral accuracy. The key engineering challenge was faithfully replicating the mixer's fixed signal processing chain and precise state-transition logic, not merely its visual appearance.

0 comments Read more at DEV Community