Skip to main content

Documentation Index

Fetch the complete documentation index at: https://reasonblocks.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

ReasonBlocks is middleware for AI agents. It sits between your agent loop and the LLM, scoring each reasoning step, detecting when the agent is struggling, and injecting targeted guidance from a live pattern store — all without changing your agent’s logic or message history.

​
What it does

On every step of the agent loop, ReasonBlocks:
  1. Scores the agent’s last reasoning step for difficulty using signals like hedging language, error density, and entity count
  2. Classifies the run into a state — FAST, NORMAL, SLOW, or SKIP — using a finite state machine with hysteresis
  3. Monitors for unhealthy patterns: infinite loops, repeated test failures, edit-revert thrashing, and more
  4. Injects targeted E-trace guidance from a pattern store into the system prompt before the next model call
  5. Routes the model call to a cheaper or more capable model based on the current FSM state
The result is an agent that catches its own stuck patterns earlier, receives relevant guidance at exactly the right time, and costs less to run by routing easy steps to faster models.

​
Who it’s for

ReasonBlocks is designed for teams building production AI agents on LangChain 1.0 or the OpenAI Agents SDK. If your agents suffer from loops, excessive tool call counts, or inconsistent task completion, ReasonBlocks gives you the observability and steering infrastructure to fix that without rewriting your agent logic.

​
Key capabilities

E-Trace injection

Pull instance-level, pattern-level, and universal guidance from a live pattern store and inject it into the system prompt at the right moment

FSM state machine

Track agent difficulty across FAST, NORMAL, SLOW, and SKIP states with configurable thresholds and hysteresis

Health monitors

Detect loops, hedging, edit-revert thrashing, and test-repeat failures automatically using a configurable monitor suite

Model routing

Route to a fast model on easy steps and a more powerful model when the agent is struggling — with zero changes to your agent code

Codebase memory

Persist and recall per-repo findings semantically across agent runs using CodebaseMemory

Token saving

Compress stale tool outputs and nudge stuck agents toward an early exit to reduce context window usage

​
Get started

Quickstart

Add ReasonBlocks to your agent in under five minutes

Installation

Install the SDK and configure your API key

How it works

Understand FSM states, E-traces, and the monitoring pipeline

LangChain guide

Step-by-step integration guide for LangChain and LangGraph agents