The spreadsheet problem
Ask most organisations for their AI inventory and you will get three answers from three teams: a wiki page of prompts, a folder of models, a list of vendor tools. None of them agree, and none of them knows that a prompt was edited in production last Tuesday.
That used to be an engineering inconvenience. It is becoming a regulatory problem. The RBI's FREE-AI report recommends an inventory of AI systems, and the RBI's 2026 draft guidance on model risk management, still a consultation draft, asks for model inventories and board-approved frameworks. The DPDP Rules ask Significant Data Fiduciaries for due diligence on algorithmic software. All of them assume you know what you run.
What every registry record carries
Identity and authorship
A unique identifier, a version and a named author for each agent; a numbered version for each flow.
History
Each saved flow version records who saved it, when, and the earlier version it was restored from, if it was.
Approvals on record
An agent carries an approval status, draft, pending, approved or rejected, with who submitted it and who reviewed it.
Restore an earlier version
A flow can be restored to an earlier saved version, and the restore is recorded.
Built for two readers
Engineering
One interface to register agents, save flow versions and record approvals.
Models are kept in Model Hub, each with its own record and evaluation history.
Audit and risk
For each agent: who submitted it, who reviewed it and its approval status. For each flow: every saved version, who saved it and when.
Where it pays off first
Banks and NBFCs
A record of the agents and flows you run, agents with their approval status and flows with their saved versions, for FREE-AI-style reviews and model-risk teams.
Insurers
Changes to claims and underwriting flows saved as versions, and agents approved, like any other product change.
Defence and government
A configuration record for the agents and flows on an accredited system, in a deployment designed to run air-gapped.
Space and critical operations
A record of the agents and flows in mission tools, versioned like flight and ground software baselines.
Start with what you already run
A Registry pilot starts with the agents and flows you build on the platform, so your team sees their records, agent approvals and saved flow versions in its own environment. Request a pilot, or read how the Orchestrator runs what the Registry records. See also the Registry page.
We do not claim that an inventory makes AI safe. It makes AI accountable, which is what lets you find out. Our alignment to the RBI's FREE-AI framework and the DPDP Act is designed to support your own assessment, not to replace it.
Frequently asked questions
What is an AI inventory?
An AI inventory is a maintained record of every AI system an organisation runs: what it is, what it is used for, who owns it, what data and models it uses, which version is live, and what evidence supports its use. Regulators including the RBI, through its FREE-AI report, recommend one.
What is the difference between a model registry and an AI registry?
A model registry tracks models. An AI registry also tracks the agents and workflows built on them. On AgentAnywhere, Registry keeps agents with their version and approval status and flows with their saved versions, and Model Hub keeps the models.
Why do prompts need version control and approval?
Because changing a prompt can change an AI system's behaviour as much as changing its code or model. In a regulated process, a prompt edit should be versioned, reviewed and approved like any other change, and recorded so it can be traced later.
Does the RBI require an AI inventory?
The RBI's FREE-AI committee report recommends that regulated entities maintain an inventory of AI systems, and the RBI's 2026 draft guidance on model risk management, issued for consultation, includes model inventories. Banks should track the final directions.
Written by
AgentAnywhere Research
The team that builds the platform and the models
AgentAnywhere Research writes about the platform, the model families and the trust layer we build and run in India. Where a figure is ours, it says what it covers; where something is a demonstration, it says so.
