- Home
- Products and Trials
- Product: Orchestrator
Deterministic,
self-healing workflows.
Regulated file movement, multi-step approval chains, and AI-assisted background work shouldn’t depend on one person remembering how the pipeline connects. Orchestrator turns that workflow into a controlled, repeatable pipeline with policy-aware routing, audit-grade logging, and approvals built in.
And it no longer just moves the data past you. Anything that flows through a workflow can be ingested into a per-tenant memory palace — chunked, searchable, and queryable as a temporal knowledge graph your agents can reach over MCP.
For operations, compliance, and engineering leaders responsible for moving regulated work through the business.
DoD SAFE, federated CUI, and approval chains — running as one pipeline.
File flows shouldn’t require a person babysitting them.
Regulated file movement — DoD SAFE handoffs, federated CUI transfers, multi-step approval chains — usually lives in a tangle of email threads, file-share folders, and one person who remembers how it all connects. Orchestrator turns that workflow into a controlled, repeatable pipeline with policy-aware routing, audit-grade logging, and the approvals built in — so the work moves and so does the evidence.
Your pipeline already sees everything. Now it remembers.
Invoices, contracts, inbound mail, scanned filings, feed data — it all crosses Orchestrator already, and historically it left no trace beyond a log line. An IngestToMemoryPalace step changes that: the text goes into a searchable store on the way through, and the relationships between what it mentions become a graph you can query later.
Each document is chunked into drawers and filed by wing (the resource series it came from — a mailbox, an SFTP feed) and room (document class or time bucket), so a search can be scoped to exactly the feed you mean. Re-ingesting the same source replaces its chunks rather than duplicating them, so a re-run of yesterday’s workflow doesn’t double your memory.
One tenant’s palaces union into a single read-only aggregate. An entity that appears in two palaces — the same vendor in your invoice feed and your contract archive — becomes one node, which is how separate feeds link up by relationship rather than by filename. Every edge stays tagged with the palace it came from, so you can always tell which feed asserted what.
Relationships are directed, typed, and valid over an interval — so the question isn’t only “who is this vendor’s account manager?” but “who was it last March, when this invoice was disputed?” Ask as of a timestamp and you get the facts that held at that instant, not today’s.
The coliseum is published as an MCP server over Streamable HTTP, so your own agents — not just Orchestrator’s — can ask it what an entity is connected to, as of a point in time. The knowledge your pipeline accumulated stops being locked inside the pipeline.
Palaces are scoped per tenant at the path level and the coliseum is built per tenant — one tenant’s memory is never searchable from another’s. For a service bureau moving several clients’ regulated data through one install, that boundary is the whole ballgame.
Ingest is queued and durable, so a slow document can’t stall a transfer, and a restart doesn’t lose it. The coliseum rebuilds on a background sweep that reads palaces read-only — no lock, no contention with ingest. Your file movement stays as fast as it was before you turned memory on.
The store underneath is BogMem on BogDB, running in-process — no extra server to operate, and the same graph engine as the rest of our stack. The workflow step depends on a stable seam rather than the backend, so swapping the store doesn’t rewrite your workflows.
Reach further out, and hold the line tighter.
Orchestrator already reaches SFTP/FTPS, SMB, OneDrive and SharePoint, IBM i, SQL and SQL CDC, HTTP/REST, mailboxes over IMAP and M365, message buses (Kafka, SQS, Azure Service Bus), and object stores — with PGP, zip, and PDF operations in the middle. Recent work extended that in two directions at once.
Remote agents
A provider-neutral resource type for agents hosted outside Orchestrator — Microsoft Foundry, Copilot Studio, or anything else you already run. Hand a step to an agent that lives in someone else’s cloud without giving up your workflow’s governance, budget ceilings, or audit trail.
Blob staging
A generic binary object-store resource for staging artifacts that remote processing needs to reach. Azure Blob Storage is the first provider; the resource type stays provider-neutral so the next store is configuration rather than a code change.
AS2 that actually verifies
“Require signed payload” and “require encrypted payload” are now enforced against real partner key material rather than inspected as a Content-Type header. We found the old check would accept an unsigned body that merely claimed to be signed, and fixed it. Ingress endpoints are rate-limited, webhook secrets fail closed, and token comparison is constant-time.
Not a chat window. A pipeline that finishes the job.
Orchestrator is an AI-agent integration platform built for background work and reliable business processing. Where a conversational UI hands you a suggestion, Orchestrator enforces output structure against predefined JSON schemas — so an autonomous workflow can survive a crash, detect its own errors, and correct them without a person watching. Its planner maps intent into a DAG of durable, stateful .NET pipeline actions backed by SQL.
Secure, reliable & highly controlled
The agent uses native framework integration rather than an opinionated black-box kernel. Sub-invocations map intent onto database-backed processing queues, so what the agent decided to do is a row you can read, not a trace you have to reconstruct.
Runaway AI spend is a configuration error, not a surprise invoice. The engine enforces daily operational budgets and a max-tokens-per-run ceiling natively; execution halts when a threshold is breached, and the snapshot of why is auditable. Infinite-loop protection tracks function calls to stop an agent recursing into its own failure.
Every execution step is durably tracked in SQL. If the server restarts mid-run, the agent resumes exactly where it left off — no lost state, no half-finished transfer nobody noticed.
Platform capabilities
- Deterministic execution with automatic JSON schema validation.
- Self-healing engine using dynamic error-correction prompts.
- Durable workflow operations via SQL-backed execution queues.
- Native AI workflow planner builds custom DAGs from natural language.
- Governance controls measuring max tokens per day and per run.
- Infinite-loop protection safeguarding against recursive LLM calls.
- Server-restart resilient state tracking for persistent workflows.
- Per-tenant memory palace ingest with a temporal knowledge graph, published over MCP.
- Remote-agent and object-store resources for work that runs outside the install.
- DPAPI-backed secret vault — credentials never sit in plaintext in a resource or trigger.
- Hundreds of tests and compliance checks go into every change.
Tiers & base pricing
Choose a base tier by operational maturity, then layer add-ons for industry-specific accelerators.
Core
Starting at $3,600/year
- Baseline workflow orchestration and scheduling.
- Core resource integrations and secure run history.
Governed Runtime
Starting at $7,900/year
- Advanced runtime controls, policy guardrails, and expanded integrations.
- Enterprise-grade governance and operational controls.
Intelligent Orchestration
Starting at $28,999/year
- AI planner, schema-enforced agent workflows, and self-healing execution.
- Token governance, bounded autonomy, and advanced intelligence features.
Add-ons
Each add-on amends your existing licence in place — no reinstall and no new activation code. Buy one at any time and re-activate to pick it up.
Governed Runtime
$5,000/year
- Enterprise connectors: Microsoft 365, SharePoint, and OneDrive.
- SSH key management, plus event-bus, cloud-storage, SaaS, and internal-API triggers.
- Raises schedule, trigger, and project ceilings.
Intelligent Orchestration
$24,000/year
- AI orchestration, planner, and execution cache.
- IBM i connectivity, plus AS2, EDI, and change-data-capture triggers.
- Daily and per-run token governance.
Requires Governed Runtime — either the add-on or the Governed/Intelligent bundle.
$24,000/yearlyTGK Memory
$5,000/year
- Per-tenant memory palace ingest as a workflow step.
- Temporal knowledge graph (coliseum) aggregated across your palaces.
- Queryable over MCP, so your agents can read it directly.
Team Collaboration
$1,200/year
- Multi-user collaboration with role-based access.
- A pool of 10 seats.
- Stacks on any tier, including Core.
Add-on features
- AI Token Capacity Packs.
- Industry Workflow Packs (Finance, National Security, Insurance).
- Premium Support and onboarding accelerators.
Specialized solution pages
Self-hosted, on your infrastructure.
What you need
- Windows Server 2025+
- Web Server (IIS) role with ASP.NET, and the WebSocket feature enabled
- .NET 10 Runtime Hosting Bundle installed
- SQL Server Standard
- Disk for the artifact, log, and memory-palace roots
Our setup scripts provision the IIS role, the WebSocket feature, the persisted Data Protection key ring, and the worker service and queue. Azure publishing is supported as an alternative to IIS.
What we bring
- .NET 10 Blazor Server web application — designer, Resource Hub, automation, dashboards
- Windows background worker — trigger adapters and the run-queue consumer
- In-process memory and graph engine — no additional server to operate
Bring Orchestrator to a real file pipeline.
Bring a real workflow — a DoD SAFE handoff, a multi-step approval chain, a recurring file move — to the 30-minute call. We’ll show how Orchestrator runs it with policy-aware routing, durable state, and the audit trail your auditor will ask for — and, if the data is worth remembering, what it looks like as a graph you can query afterwards.