open source · MIT licensed

AI Agent Tooling. Built for builders_

We're solving the real bottlenecks in AI agent pipelines — observability, memory, security, orchestration, and evaluation.

Explore tools ↓ GitHub

terminal


          $ pip install agentlens neuromem sentinel-inject

          $ pip install synapse-orchestrator evalforge

          # Zero required dependencies. Works with any LLM provider.

          $ agentlens view trace.json

          AgentLens — loaded 12 calls from trace.json

          ✓ 11 succeeded   ✗ 1 failed   312ms total

The Stack

Five tools. One mission: make AI agents observable, reliable, and fast.

agentlens

DevTools for AI agents — profile every call, find every bottleneck

Per-call latency & token accounting
Drop-in OpenAI & Anthropic wrappers
JSON trace export & CLI viewer

GitHub → Docs

neuromem

Smart context management — never lose critical memory again

Multi-factor importance scoring
Extractive & abstractive summarization
Token budget enforcement

GitHub → Docs

sentinel-inject

Prompt injection defense — screen every tool result before it hits your agent

Dual-layer: rules + LLM classification
4 sanitization modes (label/redact/escape/block)
Middleware wraps any tool in one line

GitHub → Docs

synapse-orchestrator

Parallel tool execution — auto-detect dependencies, maximize speed

Automatic dependency DAG analysis
3–10× speedup on fan-out pipelines
OpenAI & Anthropic integrations

GitHub → Docs

evalforge

Agent evaluation harness — repeatable evals, regression detection, CI-ready

6 scoring strategies (exact, fuzzy, LLM judge…)
Regression tracking across runs
HTML & JSON report generation

GitHub → Docs

5 tools

Python + TypeScript dual SDK

MIT license

Zero required deps