01
Quality tiers
Page Type Contracts
Define what each page needs to qualify for FULL, BASIC, SHELL, or SUPPRESS.
Open-source governance for agent-built publishing
Constitutional CMS is a governance framework for AI agents that build websites. It defines what agents are allowed to publish, when pages qualify, and how systems fail safely when the data boundary drifts.
From the Labs
Built inside Targeted Impressions LabsFive contracts
Each contract describes one part of the publishing system so parallel agents can work without silently undermining the site they are building together.
01
Quality tiers
Define what each page needs to qualify for FULL, BASIC, SHELL, or SUPPRESS.
02
Single-writer pipelines
Declare which stage writes which state, what it reads, and the gates that must pass first.
03
No phantom links
Constrain which page types can link to one another before a broken graph reaches production.
04
Read and write separation
Turn the shared data boundary into the inter-agent contract so drift fails safely instead of silently.
05
Live acceptance
Tie scope, ownership, and exit criteria to what the live site proves, not what merged locally.
How it works
The contract replaces hand-wavy coordination. Each role knows what it can touch, what it must not invent, and what counts as done in production.
Agents read the contract before writing code.
Write agents produce snapshot rows inside their boundary.
Read agents consume those rows and never improvise primary data.
Validation checks link integrity, tier eligibility, and live outcomes.
Write agents
Produce snapshot stateSnapshot boundary
Shared contractRead agents
Consume without improvisingProduction proof
Constitutional CMS governs SERPRadio, a live programmatic publishing system. The public case study stays at the outcome and architecture layer. The private mappings and heuristics stay private.
851
pages returning HTTP 200
54ms
median TTFB
4
AI agents working in parallel
19
merged PRs in one coordinated push
SERPRadio
The public case study focuses on publish-tier governance, link integrity, and agent coordination through contracts. It does not expose the internal route maps, threshold tables, or enrichment recipes that make the operating model proprietary.
Read the case studyOpen and proprietary
This launch publishes the governance architecture and enough proof to make it credible. It does not publish the detailed operational intelligence that turns one implementation into an advantage.
Open
Private
Guided start
Start with the overview here, then move into the repository for the full contract files, validation script, and implementation notes.
Start with the contract directory, then rename the entities and thresholds for your own domain.
The contract is read first so code generation begins from constraints instead of from improvisation.
Run structural validation locally, then verify live output against the same rules after deployment.
From the Labs
Launch surface