Agent SDK

@cognipeer/agent-sdk · Flagship · Runtime

stable

A deterministic TypeScript agent runtime for planning, safety, memory, and multi-agent orchestration. Zero-graph, message-first design with built-in summarization, guardrails, and structured tracing.

Quickstart GitHub Full docs

License · MITTypeScript · 100%Audience · Teams building custom…Fits · Custom agent orchestration

Custom agent orchestrationDeterministic workflowsTracing-rich executionBrowse recipes

Production install

$npm install @cognipeer/agent-sdk

What's in the box

Teams building custom agent logic in TypeScript and needing deterministic execution, planning, and human-in-the-loop control. Each capability is opt-in — use the parts that fit, leave the rest.

Message-First Runtime

Executes agent reasoning through a transparent, message-driven runtime with full conversational traceability.

Planning (TODO)

Structures agent workflows through managed task planning and explicit execution steps.

Pause / Continue

Persists agent state to enable workflow pausing and asynchronous continuation.

Human-in-the-Loop

Introduces approval and review checkpoints within agent decision processes.

Interrupt

Allows live agent execution to be stopped, redirected, or overridden in real time.

Summarisation

Compresses long conversational context using token-aware memory summarisation.

Guardrails

Applies runtime safety and policy controls to govern agent behaviour and outputs.

Structured Tracking

Captures detailed execution traces, tool calls, and reasoning events.

How it runs

A single deterministic loop: plan → call tools → reflect → checkpoint. No graph DSL to learn, no implicit globals. Everything you would inspect in a debugger is on the result.

Input

Message

Step 1

Plan

Reasoning budget

Step 2

Tools

Typed · approved

Step 3

Reflect

Stream + persist

Step 4

Checkpoint

Snapshot state

Output

Result

Inputs

User messages
Tool registry
Reasoning profile
Approval policies

State

state.plan
state.reflections
state.history
state.snapshot

Side effects

Tool calls
Tracing events
Approval requests
Token accounting

Quickstart

Install, configure, run. The example below is the smallest piece of code that does something useful in production.

import {
  createSmartAgent, createTool,
  createProvider, fromNativeProvider,
} from "@cognipeer/agent-sdk";
 
const model = fromNativeProvider(
  createProvider({ provider: "openai", apiKey: process.env.OPENAI_API_KEY! }),
  { model: "gpt-4o" },
);
 
const agent = createSmartAgent({
  model,
  tools: [searchKb, openTicket],
  runtimeProfile: "balanced",
  tracing: { enabled: true },
});
 
const result = await agent.invoke({
  messages: [{ role: "user", content: "Where is order #4187?" }],
});

How it compares

Against the flagship · runtime options teams most often weigh — focused on operational concerns, not feature inventories.

Capability	Agent SDK	LangChain (JS)	LangGraph	OpenAI Assistants API
Deterministic Runtime	● native	○ missing	○ missing	○ missing
Planning	● native	◐ partial	◐ partial	○ missing
Summarisation	● native	○ missing	○ missing	○ missing
Guardrails	● native	○ missing	○ missing	◐ partial
Interrupt	● native	○ missing	○ missing	○ missing
Pause / Continue	● native	○ missing	◐ partial	○ missing

Patterns that use Agent SDK

Production-tested recipes powered by this package.

recipe · 11 min

SRE Incident Response Agent

Build an SRE agent that reads Confluence runbooks, ingests them into Console RAG, searches vector knowledge during incidents, and posts Jira updates through tools.

agentsops

recipe · 11 min

Agent Development With Tracing

Develop custom agents with Agent SDK and feed execution traces into Console so teams can compare runs, inspect tool behavior, and debug workflows faster.

agentsobservability

recipe · 12 min

Autonomous AI Agents

Build agents that plan, execute tools, and deliver structured outputs with full deterministic control and observability.

agentsplanning