# Gift Graph Product Plan > Where v0.1 goes from here: the identity ladder, the three probes, and what each one measures. The concept page set the destination; this is the order of the stops. **The bet:** Personal agents are accumulating rich context about the people who use them, and there is no consent-shaped way for one agent to lend a sliver of that context to another. Every current answer is total or nothing. Gift Graph sits in the middle, where one party lends context, the other party's agent uses it only when asked, and neither side sees more than they agreed to. The couple is the smallest instance; the same primitives (pairing consent, tiered reduction, anti-inference timing, revocation, a human veto on both ends) are what a vendor's support agent and a customer's support agent need to share an outage cluster without sharing the ticket. ## The identity ladder | Version | Identity | What it proves | What a security team flags | | --- | --- | --- | --- | | **v0.1 · Seeded** (live) | Two users, static bearer tokens in Replit Secrets, rows inserted by a script | The privacy model | A static credential copied by hand | | **v0.2 · Sign-in** | Clerk handles sign-up and login; first login creates the user row; a settings page issues the MCP token; invites by email or handle | A third person can join | Tokens still live in client config | | **v0.3 · MCP OAuth** | The OAuth flow MCP clients already support; the user approves access in a browser | No one copies a token | Nothing manual left on the client | | **Enterprise · SSO and SCIM** | Sign-in through the company's identity provider (Entra, Okta); SCIM removal revokes connections and drops | The customer's controls own the door | The removal path, which ties to tombstones | Each rung removes one manual credential. Static tokens are the right prototype choice and the wrong production choice. ## The three probes Each probe is a [[Thesis - Ship Probes, Not Monoliths|probe]] in the sense of that thesis: small enough to ship, instrumented enough to learn from. | Probe | Who | Questions | Measures | | --- | --- | --- | --- | | **1 · The pair** | My partner and me, one gift cycle on v0.1 | Can he set a tier without reading documentation, and does he understand what each tier lets through? Does my agent produce a better idea at vibe tier than I would have found alone, and how does that change at category tier? Can either of us triangulate what the other dropped? | Tier comprehension · recommendation quality by tier · inference leakage under adversarial prompting, red-teamed against my own agent | | **2 · Five to ten pairs** | Friends and family who already gift each other | The same three, at a scale where the anti-inference batcher matters | The same three measures, plus how often the surprise-safe flag gets used | | **3 · The business pilot** | One vendor and one customer support team who already share context badly over email | Does tier-reduced context resolve a real ticket faster than the wall did? | Time to resolution with and without the lens; what the customer's security review asked for | v0.1 exists to find out whether four tiers are three too many. ## Status against the plan | Milestone | Where it stands | | --- | --- | | Spec and architecture diagram | Done, 2026-09-01, on the concept page | | v0.1 server, deployed | Done, 2026-10-01, built with Replit Agent, driven from Claude Code | | Privacy tests | 12 at build, plus a live fishing test and a revocation test | | Probe 1 | Ready to start once semantic matching lands; literal matching makes natural prompts miss | | v0.2 sign-in | Backlog, P1 | | Probe 2 and the batcher | Backlog, P1 | | Probe 3 | Backlog, P3 | ## Open calls - Drop scope: one connection, or the owner across all approved connections ([[Gift Graph Technical Decisions#TDR-006]]) - Reconnect after revoke: fresh start or restored access ([[Gift Graph Bug Tracker#BUG-006]]) - Whether the review screen comes before or after v0.2 sign-in; the screen is what makes tier comprehension measurable without a tools-only client **Not pursuing:** a patent. Publishing the spec is the cheaper protection and the better use of the idea.