Welcome to Hyperstruck Docs
Reasoning, memory and the APIs for building agents on Hyperstruck.
Hyperstruck is a learning and reasoning engine for AI agents. It plans, executes, reflects, and learns from every task. Use it as a standalone agent platform, integrate it with LangGraph, OpenAI Agents SDK, or CrewAI, or work directly from Claude Code and Cursor.
Recommended reading order
Start with Quickstart, then read Reasoning and Memory before diving into the API and integration pages.
Concepts
Quickstart
The fastest first-success flow: create an agent, dispatch a goal, and store a learning.
Reasoning
How a run plans, acts only on what it observed, and has to prove it succeeded before it says so.
Grounded success
Why a run refuses to report success it cannot ground in something it actually observed.
Memory
The three things an agent keeps from a finished run, and why they are kept on separate shelves.
Learning
What a learning is, how it makes your agents better over time, and what it is not.
Claims
How specific claims are remembered and checked, and how you stay in control of what your agents believe.
Episodes
The governed record of each finished run, and what it means to have one erased.
APIs
Agents API
Create agents, dispatch goals, monitor runs, and resume suspended work.
Agent configuration
Tools, guardrails, approval gates, and what an agent pays attention to when it learns.
Learnings API
Store, search, retrieve, and reinforce operational knowledge.
Plans API
Search prior workflows and review candidate learnings for reuse.
Credentials API
Bring your own LLM provider credentials with tenant defaults and agent overrides.
Claims API
Review what your agents believe, decide on claims held for review, and erase an entity.
Usage API
Inspect tenant and per-agent usage, costs, learning activity, and run history.
Integrations
Framework integrations
Use Hyperstruck with LangGraph, OpenAI Agents SDK, or CrewAI.
Claude Code and Cursor
Use Hyperstruck directly from your IDE with built-in skills.
Platform
Access control (FGA and RBAC)
Scopes, organization roles, spaces, API key permissions, and the full permissions matrix.
Sharing and visibility
When one agent's learnings can reach another, what is stripped on the way, and why claims never travel.
Security and data isolation
Tenant isolation, encrypted credentials, PII masking, and access controls.
Who this is for
- Technical buyers evaluating whether Hyperstruck is more than a prompt wrapper
- Engineers looking for clean API entry points into reasoning and learning
- Teams that want to integrate Hyperstruck into existing agent frameworks or IDE workflows