feat(WS-29): MT-0 blockers + MT-1 scaffolding; migrations 157-159 scratch-verified (H1) - #404

Merged

vjvarada merged 24 commits into

mainfrom

claude/command-center-multitenant-a30fgy

Aug 9, 2026

Commits on Aug 8, 2026

docs(saas): the tenant boundary was a deployment because the business model said so
claude
committed
70880bc
docs(saas): "one seam" was true of the request path and false of the process
claude
committed
d226230
docs(saas): 10-50 seats per customer withdraws the cost argument and answers Q5
claude
committed
636c475
docs(saas): separate the arguments that are design from the ones that are retrofit
claude
committed
7eb6138
docs(saas): pooled systems buy their safety above and below the storage, not from it
claude
committed
b799a46
docs(saas): "different customers on different versions" is four requirements, not one
claude
committed
4be19b9
docs(saas): the interesting boundary was never the database
claude
committed
8c9bb5f
docs(saas): WS-29 — the plan becomes dispatchable
claude
committed
9e40de2
docs(saas): the brain now says the same thing in every doc that mentions tenancy
claude
committed
6afa48b
fix(MT-0a): a run's credentials were readable by every other run in the process
claude
committed
722cf2a
fix(MT-0b, MT-0d): a tenant could push to our monorepo and read our provider keys
claude
committed
186342d
fix(MT-0c-1): the agents had a raw-SQL tool, and its guard was wrong both ways
claude
committed
87feb15
feat(MT-1a, MT-1b): the tenant catalog, and a migration deliberately not wired up
claude
committed
5070a9d
feat(MT-1c): the ratchets — and building them found two paths the inventory missed
claude
committed
c88d7e3
feat(MT-1c, MT-1i): the binding seam, and five cross-org predicates
claude
committed
19305f6
docs(saas): mark MT-1 ticket status — what is built, what is staged, what needs a database
claude
committed
43083fd
feat(MT-1e): a Redis client that cannot express an unprefixed key
claude
committed
fca91e4
docs(saas): MT-1e status, and the three Redis sites no ratchet can catch
claude
committed
0ba6ea1

Navigation Menu

Uh oh!

feat(WS-29): MT-0 blockers + MT-1 scaffolding; migrations 157-159 scratch-verified (H1) - #404

feat(WS-29): MT-0 blockers + MT-1 scaffolding; migrations 157-159 scratch-verified (H1)#404
vjvarada merged 24 commits into
mainfrom
claude/command-center-multitenant-a30fgy

Commits on Aug 8, 2026

docs(saas): the tenant boundary was a deployment because the business model said so

docs(saas): "one seam" was true of the request path and false of the process

docs(saas): 10-50 seats per customer withdraws the cost argument and answers Q5

docs(saas): separate the arguments that are design from the ones that are retrofit

docs(saas): pooled systems buy their safety above and below the storage, not from it

docs(saas): "different customers on different versions" is four requirements, not one

docs(saas): the interesting boundary was never the database

docs(saas): WS-29 — the plan becomes dispatchable

docs(saas): the brain now says the same thing in every doc that mentions tenancy

fix(MT-0a): a run's credentials were readable by every other run in the process

fix(MT-0b, MT-0d): a tenant could push to our monorepo and read our provider keys

fix(MT-0c-1): the agents had a raw-SQL tool, and its guard was wrong both ways

feat(MT-1a, MT-1b): the tenant catalog, and a migration deliberately not wired up

feat(MT-1c): the ratchets — and building them found two paths the inventory missed

feat(MT-1c, MT-1i): the binding seam, and five cross-org predicates

docs(saas): mark MT-1 ticket status — what is built, what is staged, what needs a database

feat(MT-1e): a Redis client that cannot express an unprefixed key

docs(saas): MT-1e status, and the three Redis sites no ratchet can catch

Commits on Aug 9, 2026

docs(saas): the execution runbook, and a 2.8x undercount corrected

fix(MT-1b): the two DB gate tests answered to a dotenv side-load, not the operator

docs(work-plan): a board its own dispatch loop can read — the 2026-08-09 consolidation

docs(WS-29): pricing answered (D18), Mem0 decided (D17), H1 result recorded

docs(specs): board records land in their owning specs; D11-era language re-swept for D15

fix(MT-0a/0b): at-limit peek precedes the first-party gate; eval follows the seam it locks