Arjun Prabhakaran
Arjun Prabhakaran is an independent systems engineer and solo creator of Deedglass, a self-hosted layer between autonomous AI agents and the systems they can change. Deedglass applies rules before tool execution, checks the actual committed change against the agent's report, and seals actions in signed records that can be verified offline.
At a Glance
- Developers and small engineering teams using autonomous coding agents
- Enterprise security, infrastructure, and compliance teams
- Hospitals and clinical teams
- Health-tech, EHR, payer, and AI-scribe vendors
- +2 more
AI Tools by Arjun Prabhakaran
(1)
Deedglass
Agent Action Verification Gateway
Discussions
No discussions yet
Be the first to start a discussion about Arjun Prabhakaran
Latest News
The AI agent blast radius formula: mathematical safety in production
Autonomous AI agent due diligence: what enterprise CISOs and VCs require
Govern Medplum FHIR agent writes with Deedglass
Nemotron 3 Ultra for AI Agent Blast Radius Checks
Products & Services
Self-hosted agent-governance and verification software. It intercepts tool calls, applies policy, supports human approval, observes actual system changes, compares them with the agent's claim, and seals signed action records with offline verification and undo support.
A local app launched by npx deedglass that opens in a browser and shows agent actions, pending approvals, rules, and locked records on the user's computer.
Healthcare-oriented use of Deedglass for AI scribes, clinical agents, FHIR systems, claims, and EHR workflows. It supports clinician approval, FHIR read-back checks, and Proven checks against PostgreSQL history where available.
SDK and tool-wrapping interfaces for governing custom tools and agent applications, including a TypeScript example using @deedglass/sdk.
Market Position
Deedglass positions itself as a verification layer rather than only a prompt firewall, trace viewer, or network gateway: it checks the underlying system's committed state after a call and creates independently verifiable records. Its comparison pages distinguish it from read-only Postgres roles and database branches, and position it as complementary to agent firewalls/gateways such as Claw Patrol and Pipelock, tracing products such as Langfuse/LangSmith, and rewind products such as Rubrik Agent Rewind. The differentiator claimed by the project is pre-execution policy plus post-execution storage-ground-truth verification for self-hosted teams.
Leadership
Founders
Arjun Prabhakaran
Independent systems engineer in India; the official Deedglass About page identifies him as the project's founder and Engineering Lead. The available source-of-truth material does not identify prior employers or companies for him.
Executive Team
Arjun Prabhakaran
Engineering Lead; Founder
Founder and principal builder of the independent Deedglass systems-engineering project in India.
Founding Story
The project was started from the conviction that autonomous agents with production credentials need evidence of physical system changes rather than conversational claims. Its stated initial vision was a deterministic storage-layer boundary: rules before execution, inspection of committed reality afterward, and cryptographically permanent records. The site says it was built in India without venture-capital committees, speculative tokens, or corporate bureaucracy, with a public open-source core and zero retention by default.
Business Model
Revenue Model
Deedglass is currently free to run, including features previously described as Pro, and is primarily self-hosted. The site says it plans to offer paid support, SLAs, and a hosted version for organizations later; nothing is for sale today.
Pricing Tiers
Every Deedglass feature is free while the project grows; it runs on the customer's own servers and Postgres.
Free self-hosted healthcare use plus a free synthetic-data pilot with setup help; the site says there are no healthcare prices yet.
Target Markets
- Developers and small engineering teams using autonomous coding agents
- Enterprise security, infrastructure, and compliance teams
- Hospitals and clinical teams
- Health-tech, EHR, payer, and AI-scribe vendors
- Organizations operating PostgreSQL and other production databases
- Teams building or governing MCP and A2A integrations
- Governing coding agents and MCP servers that can write to production databases
- Verifying PostgreSQL mutations and detecting triggers, cascades, missing WHERE clauses, or other unexpected changes
- Human-in-the-loop approval of clinical, payment, customer-data, and other sensitive writes
- Healthcare AI scribes and agents writing to FHIR/EHR systems
- FHIR and Medplum write verification and auditability
- Agent-to-agent/A2A governance and provenance