Point an AI client at Oregon government
One endpoint gives a model searchable access to Oregon statutes and rules, legislative measures, budgets and expenditures, Secretary of State audits, agency performance measures, county code, collective bargaining agreements, and the federal instruments Oregon must comply with. It is free, public, and requires no account or key.
Add it to Claude
Custom connectors are available on Free, Pro, Max, Team and Enterprise plans. Free accounts can hold one connector.
- Open Settings → Connectors in Claude.
- Click + Add custom connector.
- Paste the endpoint above as the remote MCP server URL.
- Click Add. There is nothing to configure under Advanced settings — this server takes no authentication.
On Team and Enterprise plans an owner adds it once under Organization settings → Connectors → Add → Custom → Web, and members then connect it individually from their own Connectors page.
For a client that reads a config file, such as Claude Desktop:
{
"mcpServers": {
"oregonai": { "url": "https://gateway.morficflux.com/mcp" }
}
}What to ask
Ask in plain language. The model decides which corpora to search; you do not have to know which one holds the answer. Start with “what Oregon corpora can you see?” to confirm the connection.
“What does Oregon require for wildfire hazard mitigation in new construction, and what has it spent on it?”
One question, one search, five corpora answering — the building-code standard, the bill that changed it, the appropriation, the audit, and the agency's own performance report:
| corpus | top result |
|---|---|
| executive-regulatory-frameworks | Adoption of wildfire hazard mitigation code standards for new buildings |
| oregon-legislature | SB 84 (2025R1): Exempts local governments that adopted and continue to… |
| oregon-budget | Appropriations in 2021 Regular Session House Bill 2795 |
| oregon-audits | Advisory Report: ODOT Worked Quickly to Oversee the Largest Wildfire D… |
| oregon-kpm | Department of Forestry — Annual Performance Progress Report |
“Resolve ORS 401.204 and tell me which corpora reference it.”
One statute, reached five ways. Four corpora resolve it through their sibling index into the corpus that actually holds it, so the answer is one document with its provenance intact rather than five near-duplicates:
| corpus | resolved via |
|---|---|
| executive-regulatory-frameworks | direct |
| oregon-audits | sibling:executive-regulatory-frameworks |
| oregon-kpm | sibling:executive-regulatory-frameworks |
| oregon-counties | sibling:executive-regulatory-frameworks |
| oregon-collective-bargaining | sibling:executive-regulatory-frameworks |
Three further corpora reported that they recognised no citation of that form — which the answer states, rather than leaving it as silence.
“Compare what the Department of Forestry reported achieving against what it was appropriated.”
Spending and self-reported outcomes for one agency, side by side. Answers of this shape are the reason the platform exists:
| corpus | top results |
|---|---|
| oregon-budget | Forestry, Dept of — FY2022 expenditures |
| oregon-kpm | Forestry, Department of — Annual Performance Progress Report |
| oregon-budget | Forestry, Dept of — FY2019 expenditures |
Performance figures are each agency's own report on itself — not an independent finding. The audits corpus is what disputes them.
Reading the answers
These are non-authoritative mirrors, never the official text. Every document
carries a source_url; anything that matters should be checked against it.
Ask the model to cite them — the servers instruct it to.
Search always returns something. Matching is partly semantic, so a corpus with nothing relevant still returns its nearest documents. A result is a candidate to check, not confirmation that the subject is covered.
Every response says which corpora answered. If one was slow or unreachable it is named, and the answer is marked partial. Not finding something is never the same as it not existing — ask what was actually searched.
Connecting one corpus at a time
The endpoint above fronts every corpus and is the simplest way in. Each corpus also serves its own MCP endpoint, listed on the main page — useful to pin a client to a single subject area, and unaffected by the gateway.
Everything is open source. The corpora, the shared toolkit they are built on, and the contract they all answer are at github.com/OregonAI.