Context is the upgrade.
The preview is designed for a local MCP connection to stock Apache NiFi. AI gets scoped context; your infrastructure stays yours.
NiFi confidence. Inside your environment.
Understand inherited flows, investigate with evidence, and help your team operate with confidence. Explore NiFiMCP, with deployment and hands-on support available through Federal SI.
Public preview 0.1.0rc1 is out. Live NiFi access is read-only; changes stay in your hands.
Your pipeline just got a copilot.
Preview scope and roadmap Get it on GitHubBuild an understanding of your flows for routine operations and when issues arise. Explore three example investigations, with evidence, inference, and unknowns kept distinct.
POV: 800 processors. Zero flight manual.
Synthetic example
RouteOnAttribute
Routing processor
ROUTE recorded
event 218
source.type
Absent in the sampled attributes
The sampled FlowFile took the unmatched relationship at RouteOnAttribute.
The current routing rule expects source.type. That attribute is absent in this sample. Historical rule values are unavailable, so the cause remains an inference.
Interactive illustration using synthetic evidence. No live AI session or NiFi connection. Public preview 0.1.0rc1 is out.
Practical help from Federal SI for teams building NiFi expertise, including those with limited access to specialists.
Federal SI
Work alongside your team to deploy NiFiMCP, document pipelines, review proposed changes, and preserve operational knowledge. Plan work within your approved environment, including disconnected settings.
Discuss your environmentEngineering assistance, environment-specific validation, knowledge transfer, and separately scoped support. Scope, access, and handover agreed before work starts.
Assistance installing, configuring, and validating the existing NiFiMCP offline kit for an agreed environment, with operating guidance and handover.
An engineer working alongside your operators on deployment, integration, and workflows. Remote or on-site work is scoped for each engagement.
A practical, hands-on session using representative workflows to understand NiFiMCP, interpret evidence, and establish repeatable operating practices. Standalone or as part of deployment.
Explore the open-source preview on your own. Federal SI services add hands-on assistance for your environment.
The offline installation kit does not include an AI model. A fully offline AI setup and local-model compatibility have not been tested. Any AI host or model must meet your environment’s requirements; validation is scoped separately. Restricted environments and on-site work require an agreed access and delivery plan.
Understand. Investigate. Propose. Execute with approval.
Explain inherited flows, processors, and dependencies in plain English. Give new operators a starting point for understanding pipelines.
Review status and queues during routine operations. Investigate issues through configuration and provenance, keeping evidence separate from inference.
Prepare reviewable flow drafts. Importing a draft is a separate, authorized action.
Apply a reviewed change after explicit human approval, with scoped access and an audit trail.
Live access stays read-only. Drafts do not change NiFi. Approved execution is a future capability. The public preview targets standalone NiFi 2.12.0 on Linux; clusters are excluded. Tested scope and the three blocked gates are in the GitHub README. See the project brief for scope.
Explore the project, inspect the boundaries, and help shape what comes next.
Read the project brief Read the operator field notes Download the project brief (PDF)The preview is designed for a local MCP connection to stock Apache NiFi. AI gets scoped context; your infrastructure stays yours.
Use cited findings and exported reports to document pipelines, support operator onboarding, and preserve operational knowledge. Facts, hypotheses, and unknowns stay distinct.
Released under Apache 2.0. The project and existing offline artifacts remain free. Federal SI deployment assistance, knowledge transfer, and support are optional and separately scoped.
Public preview 0.1.0rc1 is on GitHub and PyPI. Install with uv tool install nifimcp-server==0.1.0rc1, then run nifimcp demo. Tested with Apache NiFi 2.12.0, single node, certificate authentication, on Linux x86_64 with Python 3.12 and Claude Code. See the capability matrix for untested and unsupported configurations.
Tell Federal SI about your environment and what your team wants to accomplish. Ask about the Supported Air Gap Kit, engineering assistance, a Technical Exchange, or guidance on where to start. No package selection needed.
Please share only a high-level description. Keep credentials, production logs, personal data, and restricted information out of this form.