Agent SDK

Native LLM Providers — No Framework Required

Call OpenAI, Anthropic, Azure, Bedrock, Vertex, and any OpenAI-compatible API directly with fetch. Unified schema, full token tracking (input/output/cached/reasoning), and streaming built-in.

Reasoning And Reflection

Use one `reasoning` config to enable provider-native reasoning plus post-tool plain-text reflections. Reflection notes persist on `state.reflections` and stream through `reflection` events.

Runtime Profiles For Different Agent Behaviors

Use fast, balanced, deep, or research as operational presets. They are real tradeoff bundles for context budget, delegation depth, memory policy, and summarization behavior.

Planning That Fits Autonomous Work

Smart planning is aimed at agents that own multi-step work. The durable plan lives on result.state.plan instead of being trapped in transient UI events.

Summarization With Recovery Paths

Long-running agents can compact tool-heavy history without going blind. Archived outputs remain recoverable through get_tool_response when the agent needs raw evidence again.

State, Resume, And Human Control

Pause execution, snapshot state, restore later, and gate risky tools through approvals without losing the agent's working context.

Tracing, Debugging, And Evaluation

Inspect tool calls, summaries, handoffs, token drift, and partial sink failures through structured traces built for real operational debugging.

Start Here

If you are integrating the SDK for the first time, read the docs in this order:

Getting Started to get a working agent into your app fast.

Native Providers to connect to OpenAI, Anthropic, Azure, Bedrock, Vertex, or any compatible endpoint without additional dependencies.

Core Concepts to understand what actually lives in state, what is emitted as an event, and what gets summarized.

Architecture to understand the smart wrapper, the base loop, and where runtime decisions are made.

Quick Start

import { createSmartAgent, createTool, createProvider, fromNativeProvider } from "@cognipeer/agent-sdk";
import { z } from "zod";

const lookup = createTool({
  name: "lookup_owner",
  description: "Return the owner for a project code",
  schema: z.object({ code: z.enum(["ORBIT", "NOVA"]) }),
  func: async ({ code }) => ({ owner: code === "ORBIT" ? "Ada Lovelace" : "Grace Hopper" }),
});

// OpenAI, Anthropic, Azure, Bedrock, Vertex, or any compatible endpoint
const model = fromNativeProvider(
  createProvider({ provider: "openai", apiKey: process.env.OPENAI_API_KEY! }),
  { model: "gpt-4o" },
);

const agent = createSmartAgent({
  model,
  tools: [lookup],
  runtimeProfile: "balanced",
  planning: { mode: "todo" },
  limits: { maxToolCalls: 6, maxContextTokens: 12000 },
});

const result = await agent.invoke({
  messages: [{ role: "user", content: "Compare ORBIT and NOVA." }],
});
console.log(result.content);

What This Site Optimizes For

Fast product onboarding without hand-wavy architecture claims.

Clear separation between the minimal loop and the smart runtime wrapper.

Practical guidance for provider-native reasoning, reflection, and stateful execution surfaces.

Production-oriented guidance for autonomous agents, especially around planning, context pressure, and tracing.

A product-led docs surface with platform attribution kept in the footer instead of the header.

@cognipeer/agent-sdkA smart runtime for autonomous agents.