agent-stack

An agent composition framework that builds stacks and compiles specialized subagents for Claude Code

The Agent Reliability Stack: fit, guard, snap, vet, cast. One landing page for all 11 npm + 6 PyPI packages.

Public map of a production-oriented Codex and Claude Code agent stack.

MCP server: declarative network-egress firewall for agent tools. Wraps @mukundakatta/agentguard.

A framework-registry PaR trajectory designer for Hermes Agent — maps research trajectories across 12 Practice-as-Research frameworks. Part of the techne-tools organisation.

Multi-agent USDC treasury management on Arc. Six AI agents handle yield, liquidity, FX, payments, risk via Nanopayments (x402).

Cost and latency tracking for AI agent runs. Wraps your LLM call, returns a per-step + per-run breakdown.

MCP server: validate tool-call args before execution. Wraps @mukundakatta/agentvet.

MCP server: token-aware message truncation. Wraps @mukundakatta/agentfit.

OpenClaw AI Agent All-in-One Suite — Persona · Memory · Knowledge (open-upsp + memory-plus + zettelkasten)

MCP server: structured-output enforcer for any LLM. Wraps @mukundakatta/agentcast.

The agent reliability stack in one install: fit + guard + snap + vet + cast (Python ports).

Snapshot tests for AI agents — record tool-call traces, diff against baselines, fail CI on regressions. Zero deps.

Validate LLM-generated tool args before execution. Wrap your tools with a schema; throws ToolArgError with an LLM-friendly retry hint. Zero deps.

MCP server for @mukundakatta/agentbudget — token + dollar budget caps for AI agents over stdio. Drop into Claude Desktop, Cursor, Cline, Windsurf, Zed.

Token + dollar budget caps for AI agents — raises BudgetExceededError when an LLM call would push past the ceiling. Zero deps, drop into any provider SDK. Python port of @mukundakatta/agentbudget.

Network egress firewall for AI agents — declarative allowlist of domains an agent's tools can fetch, throws on violation. Zero deps.

Structured output for any LLM call — validate, retry with feedback, return typed data or throw. Zero deps. Sibling to agentsnap + agentguard.

GitHub Action for AI agent tool-call snapshot tests. Diffs current traces against baselines, posts PR comments on drift. Wraps @mukundakatta/agentsnap.

Python port of @mukundakatta/agentfit: fit your messages into the LLM context window