MeitY's e-Governance Division Picks CoRover to Build a Reusable Agentic Assistant Platform -- DigiLocker Goes First

On October 7, 2026, Storyboard18 reported that the National e-Governance Division (NeGD) under MeitY had awarded CoRover, the Bengaluru company behind BharatGPT, a contract for a "Unified AI Assistant & Bot" platform, with DigiLocker as the first service on it. No MeitY or PIB release was found; the details come from the tender record, an evaluation document dated September 24, and CoRover's own statements.

The award, as reported

• Selection was by quality-and-cost based selection (QCBS). Four companies bid; CoRover scored 100 and Kyndryl Solutions 91.24, while Cactus Technology Solutions and NEC Corporation India were disqualified at the technical stage. The value is about Rs 3.63 crore.
• CoRover's own announcement on September 30 described it as the "sole selected partner," and said a wider RFP drew more than 80 companies, with six empanelled. Those are company claims, separate from the tender record.

How the workflow is meant to work

The platform is meant to be a reusable framework other departments can deploy, not a single DigiLocker chatbot. The model is "ask, do and escalate": answer citizen questions, carry out tasks, and hand cases that need a person to staff. The first DigiLocker bot is expected to fetch and share authentic documents from source systems and manage a user's documents through conversation. Delivery is about four months, split into foundation and go-live milestones, followed by 12 months of operations and maintenance.

The tender also requires an AI governance layer: version control for prompts and models, approval and rollback, validation of tool arguments, safeguards against prompt injection and jailbreaks, human-in-the-loop workflows, and monitoring of "agent behaviour, tool calls, task completion, failures, latency and AI-related costs."

Why it matters

DigiLocker holds citizens' government-issued documents, so an agent that can fetch and share them touches sensitive data and real permissions. The reported requirements show the government buying the controls, not just the chatbot. What the coverage does not answer: how consent and document-sharing permissions will be enforced at the tool level, where data is processed, and how the agent's accuracy will be evaluated before go-live.

Analysis: the governance list is worth copying for any enterprise agent procurement -- especially tool-argument validation and per-call cost monitoring, which are often missing from vendor demos. The open question is execution: a four-month timeline for a reusable, multilingual platform with these controls is tight, and the first public evaluation results will show whether the controls were built in or added later.