SaaS Integration Research Case Study

This repository contains a scoped Product Operations research study of 50 SaaS APIs for AI-agent integration.

The deliverable is report.html. It is self-contained and can be opened locally or deployed as a static site.

Scope

The original research list contained 100 apps. This submission includes 50 completed records. Apps 51–100 are retained as explicit UNKNOWN placeholders in the raw data and are excluded from the case study.

For each completed app, the dataset records:

• Category and one-line product description
• Authentication method
• Developer-access tier
• API surface
• Buildability verdict and blocker
• Developer-documentation URL

Findings

• 40 of 50 apps are buildable today.
• 40 of 50 offer Tier 1 self-serve access.
• 80% use OAuth2.
• 40 are immediate toolkit candidates.
• 50 of 50 records were reviewed through a cited, ChatGPT-assisted verification workflow.
• The review produced 83.5% field alignment across 200 explicit pass/fail checks and led to 23 refinements.

Field alignment is an evidence-review metric, not a claim of independent human validation or ground truth.

Repository contents

• report.html — final case-study page
• final_findings.json — research records and evidence URLs
• patterns.json — aggregated findings
• verification_report.json — external verification results and citations
• main.py — research pipeline
• modern_report.py — data-driven report renderer
• analyze_patterns.py — pattern aggregation
• EXTERNAL_VERIFICATION_PROMPT.md — prompt for the external verification workflow
• import_external_verification.py — validates imported verification JSON and regenerates the report

Reproduce the report

Install the dependencies:

pip install -r requirements.txt

Regenerate the HTML from the committed data:

python generate_final_report.py

To run a new research pass, set OPENAI_API_KEY and optionally COMPOSIO_API_KEY, then run:

python run_pipeline.py

To repeat the external review workflow, follow EXTERNAL_VERIFICATION_PROMPT.md, save the output as external_verification.json, then run:

python import_external_verification.py external_verification.json

Limitations

The study is limited to 50 completed apps. Verification is ChatGPT-assisted and based on cited official documentation; unresolved or partial results are retained in the verification output rather than treated as confirmed facts.