Platform

The AgentFlow Platform

Enterprise AI agent orchestration built on Apache NiFi. Seven purpose-built processors, native clustering, complete data provenance, and a visual agent designer — everything you need to move AI agents from prototype to production.

Core Processors

Purpose-Built Agentic Processors

Seven custom NiFi processors that handle every aspect of AI agent orchestration — from LLM inference to guardrails enforcement.

🧠

LLM Inference

Multi-provider LLM integration with streaming, token tracking, and automatic retries. Supports Claude, GPT-4, Bedrock, and local models via Ollama.

successfailurerate_limit
🔧

Tool Executor

Sandboxed execution of external tools via HTTP, MCP, database, and custom protocols. Docker/gVisor isolation with timeout enforcement.

successfailuretool_error
💾

Memory Manager

Long-term and short-term memory via vector database integration. Semantic retrieval, conversation history, and knowledge base management.

successcache_hitfailure
📋

Task Planner

Decomposes complex objectives into executable sub-tasks with dependency tracking, priority assignment, and adaptive re-planning.

successneeds_inputfailure
🔀

Agent Router

Intelligent routing between agents and processors based on content analysis, capability matching, and load balancing.

successtool_callescalatefailure
👤

Human In The Loop

Configurable approval gates for high-risk actions with multi-channel notifications (UI, Slack, email, API) and SLA tracking.

approvedrejectedtimeout
🛡️

Guardrails Enforcer

Pre-LLM and post-LLM guardrails: PII detection, content policy enforcement, token budgets, iteration limits, and prohibited action boundaries.

passviolation

100's Out of Box Processors

Foundatation is ready with hundreds of production-ready processors for data acquisition, transformation, and routing — purpose-built embedded with 10's of years of best practices for turning legacy and siloed data into AI-ready pipelines.

Data Ingestion

  • GetFile / PutFile
  • GetSFTP / PutSFTP
  • ConsumeKafka / PublishKafka
  • GetHTTP / InvokeHTTP
  • ListenHTTP
  • ListS3 / FetchS3Object
  • ConsumeJMS
  • GetMongo / PutMongo

Data Transformation

  • ConvertRecord
  • JoltTransformJSON
  • TransformXml
  • ConvertAvroToJSON
  • ConvertCSVToAvro
  • ReplaceText
  • UpdateAttribute
  • SplitJson / SplitRecord

Database & Query

  • ExecuteSQL / ExecuteSQLRecord
  • PutDatabaseRecord
  • QueryDatabaseTable
  • ListDatabaseTables
  • PutSQL
  • GenerateTableFetch
  • QueryRecord
  • LookupRecord

Routing & Enrichment

  • RouteOnAttribute
  • RouteOnContent
  • ValidateRecord
  • PartitionRecord
  • MergeContent / MergeRecord
  • UpdateRecord
  • EvaluateJsonPath
  • ExtractText

Architecture

Layered Enterprise Architecture

From the visual interface to the infrastructure layer, every component is designed for production reliability.

Data Foundation

Data AcquisitionTransformationNormalizationAI-Ready Data

Interface Layer

AgentFlow 3D UINiFi CanvasREST APIFlowSync

Agent Orchestration

Agent A (Research)Agent B (Planner)Agent C (Tools)

Controller Services

LLMClientServiceVectorDBServiceToolRegistryServiceGuardrailsService

Infrastructure

LLM APIsVector DBsKafkaRedisVaultZooKeeper
AgentFlow architecture diagram

Comparison

How AgentFlow Compares

The only AI agent framework with enterprise-grade provenance, governance, and clustering built in from day one.

FeatureAgentFlowLangChainCrewAIAutoGenSem. Kernel
Data Provenance
Visual Agent Designer
Native Clustering/HA
Back-pressure/Rate Control
Human-in-the-Loop~~
Deployment & Mgmt Library
Tool Sandboxing~
PII Detection/Redaction
Cost/Token Budgets~
Production Monitoring~~~
Enterprise Auth (LDAP/SSO)
10+ Year Track Record
Full support~ Partial Not available

Ready to See It In Action?

Schedule a live demo with our engineering team. See how AgentFlow transforms AI agent development from prototype scripts to production-grade orchestration.