Status: public preview 0.1.0rc1 is out. Get the source on GitHub or install nifimcp-server from PyPI. The website demo uses synthetic evidence and does not connect to NiFi.
The preview and the ambition
- Understand: preview scope
- Read scoped flows, configuration, and status to explain how a flow works.
- Investigate: preview scope
- Inspect queues, alerts, and available provenance. Separate observations, inferences, and unknowns.
- Propose: preview scope
- Prepare reviewable flow drafts. Drafting does not change a running NiFi instance; import and promotion are separate, authorized actions.
- Execute with approval: roadmap
- Apply reviewed changes with explicit approval and scoped access. This is not part of the preview.
Compatibility boundaries
The release candidate targets standalone Apache NiFi 2.12.0 on Linux. Clusters are excluded. macOS, Windows, ARM, other NiFi versions, token-based login, and clients beyond Claude Code are not claimed as validated.
Live NiFi access is read-only. An offline installation kit does not establish that an entire AI deployment is approved for an air-gapped environment. Model, infrastructure, access, and software-import requirements need their own review.
The GitHub README provides installation instructions, tested scope, and the three blocked gates. See the capability matrix for untested and unsupported configurations. No AI-accuracy percentage, production customer result, clearance, or certification is claimed here.
Open source, with optional engineering
The release is licensed under Apache 2.0. Engineering services are separately scoped: investigation, agreed implementation, validation, and operator handover. Remote and on-site delivery depend on the engagement’s access plan.
Independent project
Apache, Apache NiFi, and NiFi are trademarks of the Apache Software Foundation. NiFiMCP is not affiliated with or endorsed by the Foundation.
Download the project brief (PDF)