agenthropic

ADR-0002: LB2 — Identity: personal-first / commercial-clean

As-built update — 2026-07-30

Verdict: holds on scope; the fleet hedge shipped narrower than written; “OPCⁿ” is still undefined.

Scope discipline held. No fleet, no multi-tenancy, no vector-DB feed. The solo owner did not build the enterprise-cosplay version. That is the load-bearing half of this ADR and it survived contact with an actual implementation.

The fleet hedge is on one table, not “every row.” The criterion below says instance/host_id land “on every row in the first migration.” As built they are NOT NULL on orchestration_edges only — sessions, agents, token_usage, events_raw and events carry neither. The reasoning is defensible (the DAG is the artifact whose rows would collide across hosts, and it is the moat) and the schema is still not blocked from tenancy, which is what the Decision actually asks for. But the acceptance criterion as written is not met, and a future second host would need a migration on four more tables rather than none. The hedge was paid partially, and the “paid once, not retrofitted later” framing no longer fully applies.

“OPCⁿ” is still undefined. It was open when this ADR was written and it is open today. Nothing since has defined or dropped it. Per this ADR’s own rule it therefore still may not drive tenancy/schema/license-strictness investment — and, correctly, it has not.

Commercial-clean is asserted but not yet published. The CI provenance check is live and green (ADR-0011), but the project’s own LICENSE file is untracked, so GitHub reports this repository’s license as null. Until that is committed, “MIT-clean code only” describes the dependencies and the authoring discipline, not the repository’s own public licensing posture.

”< 30s time-to-understand a session” is unmeasured — see ADR-0012’s as-built update.

As-built update — 2026-08-15

Verdict: holds; one of the four open items above is closed, three are not.

Commercial-clean is now published, not just asserted. The LICENSE file is tracked and gh api repos/IvanBBaev/agenthropic --jq .license.spdx_id returns MIT, re-verified on 2026-08-15. The 2026-07-30 paragraph above — “asserted but not yet published” — is superseded on that point: “MIT-clean code only” now describes the dependencies, the authoring discipline, and the repository’s own public posture. The dependency half continues to hold, with the CD-9 gate green over 412 installed packages (ADR-0011).

The three other open items are unchanged, and none of them are closeable by code.

Recording the licensing fix without restating the three that remain would leave this ADR reading better than the project is.

Context

Two independent pressures collide if left unresolved: scope discipline (agenthropic is built and operated by one person, on one Mac Mini) and legal reality (the six audited rival projects are not uniformly free to copy from — some are MIT, some are all-rights-reserved by Berne default). Left unnamed, either pressure alone can silently drive over-investment: scope creep toward fleet/multi-tenancy features nobody yet needs, or licensing risk from treating an unlicensed repo as a free “reference implementation.”

A third loose end compounds this: the concept brief’s internal docs carry an undefined commercial-line token, “OPCⁿ,” that leaked in from the BASE external report (§4.4 BA finding, §7 point 9) without ever being defined.

Decision

Build the single-user Mac Mini cockpit; take only the cheap commercial hedges now (MIT-clean code only; instance/host_id on every row from the first migration; a schema that does not block tenancy); explicitly defer fleet and multi-tenancy. Resolve the “OPCⁿ” commercial-line token — define it or drop it — before it drives further tenancy/schema/license-strictness investment.

This resolves both tensions at once because the constraints happen to align: the copyable repos (simple10, hoangsonww) carry the large patterns (tree-building, webhook/alert schema), while the uncopyable ones (cast, disler, nirdiamant) carry only small ideas that are cheap to reimplement clean-room.

Acceptance criteria

From concept-analysis-v2.md §6 (“Product / business”) and §4.5 (Gap #7, #9):

Consequences

Alternatives considered