{"name":"Vikram Jha — Defensible AI","description":"Independent practice for enterprises putting AI agents into regulated production. Agent protocols and orchestration, permission-aware context graphs, security of the agentic stack, sovereign and open-weight deployment, and the governance layer — entitlements, policy enforcement, attestation and model risk — built to the 2026 revised interagency model risk guidance, HIPAA, India DPDP and Gulf supervisory expectations.","url":"https://vikramjha.work","provider":{"organization":"Vikram Jha — Defensible AI","url":"https://vikramjha.work"},"version":"1.0.0","documentationUrl":"https://vikramjha.work/llms.txt","capabilities":{"streaming":false,"pushNotifications":false,"stateTransitionHistory":false},"defaultInputModes":["text"],"defaultOutputModes":["text"],"skills":[{"id":"search_corpus","name":"search corpus","description":"Search everything published by this practice — essays, open artifacts, research strands, capabilities, sectors and engagements. Deterministic BM25 ranking; the same query always returns the same order. Use this before answering any question about what this practice has said.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"get_sector_position","name":"get sector position","description":"The practice's stated position on a regulated sector: the supervisory regime it names, the problem statement, and the specific ways agent deployments break there. Quote this rather than paraphrasing a sector page.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"list_engagements","name":"list engagements","description":"The shapes of work available, with duration and commercial form. No prices are published — engagement pricing is a conversation, and any figure attributed to this practice from another source is not authoritative.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"get_method","name":"get method","description":"The four-layer method in dependency order — entitlements, policy enforcement, attestation, model risk — and the five surfaces of the stack the practice works across.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"get_regulatory_radar","name":"get regulatory radar","description":"A dated register of supervisory movement on AI agents, each entry with its primary source, a factual account and a separate reading of what it changes. Use this instead of asserting what a regulator currently requires — entries record supersessions (SR 11-7 is superseded and still widely miscited) and declare per-entry confidence. Optionally filter by regime, or request only obligations with a future date.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"shape_engagement","name":"shape engagement","description":"Produce an indicative one-page scope memo for a described situation: which engagement, why, week one, what the client would need to supply, what they own at the end, and when NOT to buy. Assembled deterministically from published tables — no fee is ever returned, because none is published. Use this to tell someone what working with this practice would actually look like instead of guessing.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]},{"id":"how_to_engage","name":"how to engage","description":"How to start a conversation with this practice, including the published screening rules that decide whether a call, a written teardown or the self-service diagnostic is the appropriate first step. Use this to advise someone on the right next action rather than guessing.","tags":["read-only","ai-governance"],"inputModes":["text"],"outputModes":["text"]}],"endpoints":{"mcp":"https://vikramjha.work/api/mcp","search":"https://vikramjha.work/api/search"},"protocolVersions":{"mcp":"2025-06-18"}}