Introduction to Agentic

Introduction to Agentic

Agentic is a toolkit for AI agent orchestration. It gives you the primitives to spawn, coordinate, and manage AI agents across TypeScript and Python without locking you into a single framework or provider.

The Problem

AI agent tooling is fragmented:

• Framework lock-in — CrewAI, LangGraph, and Strands each have different APIs for the same concepts. Switching frameworks means rewriting everything.
• Token complexity — Managing GitHub tokens across multiple organizations is manual and error-prone. One wrong token breaks CI, PRs, or agent spawns.
• No fleet coordination — Spawning one agent is easy. Coordinating ten across different repos, monitoring their progress, and handling failures? That requires infrastructure.
• Scattered tools — Your triage system is separate from your fleet manager which is separate from your 3D pipeline. No shared patterns.

The Solution

Agentic is a focused set of packages that solve these problems independently or together:

@jbcom/agentic

TypeScript

The orchestration layer. Spawn AI agent fleets, route tokens per organization, execute tasks in sandboxed Docker containers, and hand off work between agents with full context preservation.

agentic fleet spawn "my-org/api" "Fix failing CI" --auto-pr
agentic sandbox run "Security audit" --workspace . --timeout 900

@jbcom/agentic-triage

TypeScript

Portable triage primitives built on the Vercel AI SDK. Issue management, PR review, sprint planning — as composable tools that work with any AI provider and any project tracker (GitHub, Jira, Linear, Beads).

import { anthropic } from '@ai-sdk/anthropic';
import { generateText } from 'ai';
import { getTriageTools } from '@jbcom/agentic-triage';

const result = await generateText({
  model: anthropic('claude-sonnet-4-20250514'),
  tools: getTriageTools(),
  prompt: 'Triage all open critical bugs',
});

agentic-crew

Python

Framework-agnostic crew orchestration. Define agents and tasks in YAML, run them on CrewAI, LangGraph, or Strands — whichever is installed. Includes a universal CLI runner for aider, claude-code, ollama, and other single-agent tools.

from agentic_crew import run_crew

# Auto-detects best framework at runtime
result = run_crew("my-app", "reviewer", inputs={"pr": 42})

meshy-content-generator

TypeScript

Declarative 3D asset pipelines for Meshy AI. Define text-to-3D, rigging, and animation pipelines in JSON. Built-in 3D preview, VCR recording for CI, and manifest-based state that survives failures.

content-gen run character ./assets/characters/hero

Key Principles

Provider agnostic — Every package works with multiple AI providers. Anthropic, OpenAI, Google, Mistral, Ollama, Azure. Swap providers by changing a config value.

Declarative over imperative — Crews in YAML. Pipelines in JSON. Fleet config in JSON. Token routing from environment variables. Configuration lives in version control.

Security first — Token sanitization everywhere. SHA-pinned GitHub Actions. OIDC publishing. Sandboxed execution with Docker resource limits. Zero hardcoded credentials.

Polyglot — TypeScript for orchestration and web tooling. Python for AI framework integration and crew runtime adapters. Each language where it is strongest.

Who Is This For?

You are…	Agentic gives you…
A developer managing multiple GitHub orgs	Automatic token routing — one config, all orgs
A team standardizing AI workflows	Framework-agnostic crews — define once, run anywhere
A platform engineer coordinating agents	Fleet management — spawn, monitor, coordinate, hand off
A support or QA team drowning in backlog	Structured triage, review automation, and issue routing
A 3D artist building asset pipelines	Declarative Meshy pipelines with manifest state

Next Steps

1. Quick Start — Install and spawn your first agent in 5 minutes
2. Configuration — Set up tokens, providers, and fleet defaults
3. Packages — Deep dive into each package
4. Guides — Learn orchestration patterns and best practices