MCP server for per-agent cost attribution, budget management, and spend analytics across AI agent fleets

Unified AI coding agent cost tracking + budget guardrails — Claude Code / Cursor / Codex / MCP logs to one cost report

AI agent cost guardrails. Budget enforcement per user/session. Block, degrade, or warn before LLM calls exceed limits.

Drop-in agent-cost telemetry SDK: auto-instruments the Anthropic + OpenAI SDKs (ESM, via import-in-the-middle), maps per-call token usage to the Proxon-native agent-cost event, and fire-and-forget POSTs it to the ingest tier (POST /v1/agent-cost/events).

Core SDK for NeuraMeter — AI agent cost tracking

Metrx MCP Server — AI agent cost intelligence tools for LLM agents. Track spend, optimize models, manage budgets, detect waste, and prove ROI.

Local preflight cost estimates for coding-agent runs

Local-first MCP server for Claude Code cost analysis, budgets, forecasts, anomalies, and subagent-aware optimization from local JSONL session logs.

Freeze the model, evolve the harness. Two measured applications: (1) SWE-bench code-repair — conformant GLM->Opus empty-patch cascade resolves 51.3% Lite (n=300) and 55.6% Verified (278/500, Wilson 95% CI [51.2, 59.9], official swebench gold eval, no gold

Static analysis CLI that finds deterministic LLM-agent cost-exposure bugs (unbounded retry loops, unbounded while(true) agents, leaked system prompts) and computes a worst-case dollar exposure ceiling from real, dated model pricing.

Budget limits and cost guardrails for AI agents. Prevents runaway API spend with hard budget enforcement, circuit breakers, and per-agent cost tracking.

The agentic meta-harness — freeze the model, evolve the harness. An open runtime that routes each query to the cost-optimal model, evolves its own harness (planner/context/reviewer/retry/tool/memory/score policy) and autonomously repairs code, then orches

AWS SDK for JavaScript Cost Explorer Client for Node.js, Browser and React Native

Limit the cost of a GraphQL Query.

MetaHarness — mint a custom AI agent harness from any repo. Browser Studio + `npx metaharness` CLI. Runs on Claude Code, Codex, pi.dev, Hermes, OpenClaw, RVM, Prime Agent.

Fine-tune cheap open-source LLMs (GLM, Qwen, DeepSeek) on your AI coding agent's successful runs with LoRA (SFT + DPO) so your model cascade escalates to expensive frontier models (GPT, Claude) less often — cutting cost-per-resolved. Turns run history int

In-process spawn subagent backend: runs a fresh child agent on ctx.agents

Turn a function into an `http.Agent` instance

A verifiable self-improvement loop for agent harnesses. Freeze the model. Evolve the harness. Promote only what proves lift.

AI DevOps Framework - AI-assisted development workflows, code quality, and deployment automation