Skip to main content
Orchestrator · 2026 Q1

Multi-agent AI workflows.
At enterprise scale.

Orchestrator is the enterprise platform for designing, executing, and operating multi-agent AI workflows. A drag-and-drop visual builder, a distributed runtime, an MCP-based plugin system, and a one-shot infrastructure-as-code deploy to AWS, Azure, GovCloud, or on-prem Kubernetes.

Orchestrator template detail page for the RFI Response Pipeline — showing the multi-node workflow graph, version history, success rate, and run history.

MCP

Native plugin protocol

REST

Any HTTP-speaking system

Any

Frontier or open-source LLM

SDK

Build your own plugins

See it in action

Real workflows, real metrics, real runs.

Every screenshot below is from a live Orchestrator instance — not a mockup. Project dashboards, the operations console, and a single workflow run with its full execution timeline.

Project dashboard with KPI cards (185 runs, 87% success), an execution-trends bar chart, recent runs, top templates, and recent failures.
Project dashboard — KPIs, execution trends, and recent activity in one view.
Workflow runs operations console showing 185 runs with status, template, duration, started timestamp, and run ID columns; filters at top.
Operations console — 185 runs at 87% success, filterable by template, status, and date.
Single workflow run detail page showing 5 completed nodes with real timing data — Intelligence Intake (26ms), Fetch Unprocessed Reports (214ms), Classify & Prioritize (15.8s), Route by Critical Count (28ms), Generate Routine Summary (18.1s).
Run detail — every node, every timing, every input and output.

Headline capabilities

The platform behind production AI.

Visual workflow builder

Drag-and-drop with live variable autocomplete, design-time graph validation (reachability, dangling edges, placeholder scan, branch-schema consistency), and one-click run-the-draft.

Distributed runtime

Cooperative cancellation, scheduled runs (cron + interval), human-in-the-loop approval gates. Long-running pipelines fan out and rejoin reliably.

Every major LLM in the box

Native integrations with OpenAI, Azure OpenAI, AWS Bedrock (frontier and open-source models in one integration), Google AI (Gemini), xAI (Grok), and any OpenAI- or Anthropic-compatible endpoint — covering every frontier and open-source model. Per-role routing, encrypted at rest, no lock-in.

Multi-agent orchestration with sharding

Planner → workers (parallel, sharded by token budget) → finalizer pattern. Structured output enforcement via JSON Schema with automatic retry.

First-party plugins + customer SDK

Foundational nodes plus integrations for ServiceNow, Web Search, Video Analysis, Bullhorn, RFI Generator, Mission Ops, FSx Migration, and more. Connect any MCP server or REST endpoint. Customer plugins via approved upload.

Self-describing workflows

Every workflow exposes input/output JSON Schemas, fields, globals, and example payloads so any external agent can discover and call them.

Foundational node catalog

Foundational nodes. Every workflow uses them.

triggerWorkflow entry point — declares the input schema
agentLLM call with optional tool calls, structured output, web search
conditionalRoutes flow via typed expressions (leaf or compound), labelled branches, per-branch output schemas
webhookHTTP call (any method, templated URL/headers/body)
transformBuild a JSON object from variable references
mergeCombine multiple incoming paths (object_merge / array_concat / collect)
validateDeclarative checks (required, type, enum, regex, min/max); fail or route
delayWait relative or absolute time, then continue
foreach / joinFan out across an array (serial or parallel); aggregate via collect / flatten / merge
orchestratorPlanner → workers (parallel, sharded) → finalizer
approvalPause and wait for a human via the approve API
document_parseObject-store document → markdown
document_generateMarkdown → PDF / DOCX artifact
customSandboxed code that takes (config, context) and returns a dict

LLM provider flexibility

Every major LLM. Per-role routing.

Six native integration paths unlock every frontier model and every open-source model on the market. The connection abstraction is project-scoped and encrypted at rest. The same physical credential can be wrapped in multiple connections to route per role: planner = strong model, worker = fast model, finalizer = strong again.

OpenAI

Full frontier lineup · native web search

Azure OpenAI

Vision deployments · GovCloud-eligible

AWS Bedrock

Frontier and open-source models in one integration · native streaming

Google AI (Gemini)

Native search grounding · multimodal input

xAI (Grok)

Native live search

OpenAI-compatible

Any open-source or self-hosted model

Architecture

Four services. One deployment.

Customer-hosted microservices with a single shared deployment. Production runs on AWS, Azure, GovCloud, or on-prem Kubernetes — the same deployment runs on a laptop.

  • Web

    Dashboard, workflow builder, in-app docs, embedded assistant

  • API

    REST surface, authenticated workflow dispatch, project / integration / LLM management

  • Worker

    Workflow node execution, agent runtime, tool integrations, video pipeline, document parse/generate

  • Scheduler

    Cron + interval triggers; safe to scale horizontally

Data plane

Production-grade plumbing.

  • Relational store + vector search

    System of record. Vector ANN for video frame search. Full-text search for transcript chunks.

  • In-memory cache

    Fail-open everywhere.

  • Managed message broker

    Powers the orchestrator and foreach fan-out / rejoin pattern.

  • Object storage

    S3, Azure Blob, GCS, MinIO, or local FS. Raw videos, document originals, generated PDF / DOCX.

Use cases

From RFI in to PDF out — in one workflow run.

RFI / RFP response automation

RFI Generator blueprint produces PDF + DOCX from a company KB and an RFI PDF in one run.

Mission briefs & intel triage

Mission Ops plugin ships three blueprints for DoD / IC mission planning workflows.

ITSM ticket triage

ServiceNow plugin for KB-grounded incident automation with 2 prebuilt ITSM blueprints.

Recruiter copilot

Bullhorn plugin for note triage, recruiter activity rollups, smart search.

Storage migration

FSx Migration plugin for SMB → AWS FSx with verification + reporting.

Video search & analysis

Multimodal embeddings + vector search. Index cheap, query smart.

Enterprise hardening, shipped today.

Production-ready security from day one. Not a roadmap promise.

  • Append-only audit log with CSV export and configurable retention
  • Scoped, expiring API keys
  • Signed outbound webhooks with SSRF guard
  • Token-bucket rate limiting + idempotency-key support
  • Workload-identity-bound cloud access · secrets fetched at process start
  • Hardened containers, network-policy templates
  • Site-admin RBAC + scoped, in-process embedded-assistant contract

Technical specifications

Architecture
Customer-hosted microservices
Workflow engine
Distributed task runtime with cooperative cancellation and approval gates
Database
Managed relational store with vector and full-text search
Cache / Pub-Sub
In-memory store with fail-open semantics
Storage
S3, Azure Blob, GCS, MinIO, or local filesystem
Deployment
Infrastructure-as-code on AWS, Azure, GovCloud, or on-prem Kubernetes
Local dev
Same deployment runs on a laptop and on a production cluster

See Orchestrator in action.

30 minutes. Live walkthrough of the visual builder, plugin catalog, LLM provider routing, and a real RFI pipeline running end-to-end.