agentsforge

AgentForge

Complete AI agent infrastructure in one command — works with any Python agent framework.

Persistent memory, health monitoring, and a live dashboard. Drop it into your existing agent stack or run it standalone.

pip install agentsforge
agentforge init
agentforge start

Dashboard: http://localhost:7842

---

Supported Platforms

Default: standalone — works everywhere.

---

Quick Start

Standalone (any agent framework)

pip install agentsforge
agentforge init
agentforge start

OpenClaw

pip install agentsforge
agentforge init --platform openclaw
agentforge start

LangChain

pip install agentsforge
agentforge init --platform langchain --workspace ./my_agent
agentforge start

Docker

git clone https://github.com/JakebotLabs/agentforge
cd agentforge
docker-compose up -d

---

Commands

---

What's Included

• Persistent Memory — Three-layer memory system (Markdown + ChromaDB vector search + NetworkX knowledge graph). Agents remember context across sessions without custom infrastructure.
• Agent HealthKit — Runtime health monitoring. Detects context bloat, resource pressure, model failures. Observe-only by default.
• Dashboard — Web UI at localhost:7842. Search memory, view health status, inspect agent state.

---

Configuration

Config at ~/.agentforge/agentforge.yml:

version: "1"
platform: standalone
workspace: ~/.agentforge/data
memory:
  enabled: true
healthkit:
  enabled: true
  mode: observe
dashboard:
  enabled: true
  port: 7842

---

Why AgentForge?

Most agent frameworks solve the intelligence problem. Almost none solve the operations problem:

• Agents forget everything between sessions
• No visibility into what's happening at runtime
• No standard way to detect drift, bloat, or failures
• Every team builds their own monitoring from scratch

AgentForge is the ops layer. You bring the agent; we handle memory, health, and observability.

---

Python API

from agentforge import AgentForge

af = AgentForge(platform="standalone")
af.init()

# Store memory
af.memory.store("user prefers concise replies")

# Retrieve relevant context
context = af.memory.search("user preferences", n_results=3)

# Health check
health = af.healthkit.check()
print(health.grade)  # A, B, C, D, F

---

License

MIT — JakebotLabs