# Loomwell

> Monetizing authorized finance AI agents’ traces and logs through reinforcement learning environments licensed to frontier AI labs.

## Current finance direction

The public private-credit demo is one fictional, authored development world. Visitors inspect five source documents, build a structured credit memo and use the implemented deterministic grader. Mechanical consistency, authored failure detection and authored policy judgment are scored separately. No model runs or training takes place in the browser demo. It uses no supplier records and is not a held-out benchmark or calibrated expert judgment.

- [Finance environments](https://getloomwell.com/demos): Interactive structured tasks across ten finance segments, plus a reconciliation workbench; inspectable deterministic grades and authored examples.
- [Connected accounting pipeline](https://getloomwell.com/demo/pipeline): Fictional supplier intake, correction review, package building and buyer evaluation.
- [Interactive private-credit demo](https://getloomwell.com/demo/private-credit): Scripted agent replay, senior correction, and fictional Northwind Logistics memo grading.
- [For Frontier Labs](https://getloomwell.com/for-ai-teams): Environment licensing, package scope, integration and acceptance requirements.
- [Evaluation conversation](https://getloomwell.com/pilots/buyer): Contact form for business context only.
- [For Finance Agent Companies](https://getloomwell.com/for-agent-operators): Partnership fit, approved source scope, release review and licensing revenue terms.
- [Finance contributor discovery](https://getloomwell.com/finance-contributors.json): Contact availability and proposed payment conditions; no source upload or wallet payouts.
- [Contributor conversation](https://getloomwell.com/pilots/supplier): Business summary only; no source intake.
- [About Loomwell](https://getloomwell.com/about): Implemented demonstration and proposed supply path.
- [FAQ](https://getloomwell.com/faq): Synthetic labels, current limits and pilot scope.
- [Privacy](https://getloomwell.com/privacy): Application information handling.

Applications are open. Apply with business context only and explicit contact consent. Do not send client records, borrower documents, credentials, confidential deal information or MNPI. Finance supplier intake, licensed finance releases, expert calibration, buyer demand and training gains remain unverified or proposed.

## Historical ecommerce interfaces

These retained endpoints describe the earlier ecommerce pilot, not finance intake or finance traction.
- [Agent contributor guide](https://getloomwell.com/agents): Legacy ecommerce eligibility and trace requirements.
- [Pilot discovery JSON](https://getloomwell.com/agents.json): Legacy ecommerce machine-readable contract.
- [Application schema](https://getloomwell.com/schemas/pilot-application.json): Existing contact roles and bounded fields.
- [Trace schema](https://getloomwell.com/schemas/trace.json): Ecommerce trace model; not a finance upload endpoint.
- [Synthetic format sample](https://getloomwell.com/samples/ecommerce-synthetic-v1.zip): Three fictional ecommerce examples; no customer data.
There is no public trace upload or funded task-claiming API. An agent needs operator authorization and contact consent before applying. Uploading or passing review alone does not trigger payment; there are no guaranteed earnings.

## Historical knowledge-service tools

The existing MCP integration publishes reviewed API claims, citations and verification dates. It does not operate the finance environment or provide finance source access. Full knowledge records and account details require authorized API access.
- [Legacy API coverage](https://getloomwell.com/coverage)
- [Legacy developer FAQ](https://getloomwell.com/faq/getting-started)

## Historical OpenAI workflow guides

- [Set up the OpenAI Python SDK](https://getloomwell.com/guides/openai/python-setup): Create an isolated Python environment before making your first API request.
- [Authenticate OpenAI API requests securely](https://getloomwell.com/guides/openai/authentication): Keep the API key on your server and send it as a bearer credential.
- [Make your first OpenAI Responses API request](https://getloomwell.com/guides/openai/first-response): Send a small text request and display the generated answer in Python.
- [Read text from a Responses API result](https://getloomwell.com/guides/openai/read-response-output): Use the SDK text helper and inspect output items when your application needs more than text.
- [Stream OpenAI Responses API text in Python](https://getloomwell.com/guides/openai/streaming): Display text deltas while tracking whether the response actually completes.
- [Continue a conversation with previous_response_id](https://getloomwell.com/guides/openai/conversation-state): Connect a follow-up request to an earlier response without rebuilding the entire conversation yourself.
- [Parse structured OpenAI output with Pydantic](https://getloomwell.com/guides/openai/structured-outputs): Define an output shape for application data instead of extracting fields from prose.
- [Return a function result to the Responses API](https://getloomwell.com/guides/openai/function-calling): Complete the tool-call loop with a local demonstration function.
- [Fix OpenAI API 401 authentication errors](https://getloomwell.com/guides/openai/401-authentication-errors): Check credentials and project access before retrying a rejected request.
- [Diagnose OpenAI 429 rate limits and exhausted credits](https://getloomwell.com/guides/openai/429-rate-limits-and-credits): Use the error body to distinguish a pacing problem from a billing or spending limit.
- [Robots policy](https://getloomwell.com/robots.txt)
- [Sitemap](https://getloomwell.com/sitemap.xml)