Remember when MCP launched? The buzz was that it was “the new standard to access information.” Then the fine print came out—and we realized it was basically a JSON wrapper. Helpful for motion, yes, but not for stability.
Here’s the reality: the more tools you give an LLM, the more chaos creeps in. And chaos is a clear sign that standardization is needed.
Over the past weeks (and still going on), we’ve iterated on using natural language to drive operations in an ECM (Flowerdocs in our case, but easily mappable to any). The results are striking: hundreds of clicks replaced, pre-wired developments avoided. But a recurring question remains: what is the right granularity of tools? For instance—when describing a set of documents that respect some constraints, should you have a label metadata tool, that calls also a metadata tool? These small but crucial decisions define whether things work smoothly or not. And mutualizing thoughts of several people with different perspectives could benefit users in the end here!
Unifying experiences across system is what we have done with ARender the content viewer in the past, and that’s why we are proposing OpenCP (Open Content & Process, or Open Content Plaform): an open source initiative to standardize a core toolkit of 50–100 tools for content management and process operations. Each tool would have:
The goal is not to replace vendor innovation, but to create a common denominator that LLMs, ECMs, and integrators can all rely on. Even partial adoption brings value—and as LLMs train on these shared tools, they’ll get better at orchestrating them reliably.
So maybe the “new CMIS” isn’t MCP at all. It’s OpenCP: a collaborative, predefined set of tools for content and process management in the AI era—ambitious enough to extend into business workflows too. What do you think?
If this resonates with you, contact me and let’s discuss how to start this initiative. Let’s make it an open source standard that accelerates everyone!