OSA Procurement Guard

io.github.omarfffPublished by GitHub account omarfffv0.2.2Updated Oct 24, 2026

Pre-purchase trust, pricing and budget decisions for AI agents buying APIs and MCP tools.

VerifiedStreamable HTTPWeb executableFinanceBusiness & Commerce

Overview

AI-generated overview

A remote MCP server that gives an assistant a pre-purchase ALLOW, HOLD or DENY decision on buying an API or MCP tool, based on budget policy and trust evidence.

What you can do
Request a procurement decision before buying an API or MCP tool, returning ALLOW, HOLD or DENY based on budget policy, endpoint-scoped trust evidence, price history and optional cheaper equivalents.
When to use it
Use it when an assistant is about to purchase or route spend to an API or MCP tool and you want a policy-based go/no-go check before committing money.
Requirements
A remote endpoint reached over streamable HTTP. A required OSA API key supplied in the X-API-Key header; the provider says 20 free decision calls are available from its website.
Before you install
The API key is a secret sent in the X-API-Key header. The decision is advisory and depends on the provider's own trust and pricing data, so treat ALLOW as a recommendation, not a guarantee. The README also describes paid products and machine-payment paths, though the single exposed tool only returns a decision.

Installation

In SourceWeft

  1. Open OSA Procurement Guard in the dashboard and add it to a workspace.
  2. Connect X-API-Key credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.omarfff%2Fosa-procurement-guard/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "osa-procurement-guard": {
      "type": "http",
      "url": "https://pkctqxeydfuiupadaoov.supabase.co/functions/v1/procurement-mcp",
      "headers": {
        "X-API-Key": "<YOUR_API_KEY>"
      }
    }
  }
}

README

OSA

OSA is a machine-commerce trust and procurement system for evaluating agent/API endpoints before purchase or routing spend.

Core product

OSA combines live endpoint verification with historical evidence and produces a TrustScore plus confidence and reason codes.

  • GET /best — rank matching endpoints by trust, intent and optional max price.
  • GET /score — live-verify a registered endpoint.
  • GET /history — return historical snapshots.
  • POST /ingest — authenticated registry ingestion.
  • POST /sources/mcp and POST /sources/bazaar — ingest MCP/Bazaar-shaped records.
  • /.well-known/osa.json — machine-readable product metadata.

TrustScore currently considers uptime, latency, price stability, schema stability, payment stability and transaction evidence. Confidence is reported separately.

Architecture

OSA uses a split-plane design:

  • Product/Data/Commercial plane — procurement/MCP distribution, pricing and benchmark history, endpoint observations, product usage, leads/deals and payment reconciliation.
  • Control/Execution/Security plane — VPS agents, approvals, task policy, audit/control alerts, SSH bridge and certificate-only RAW SSH execution.
  • GitHub — canonical source, CI/security scanning and controlled deployment workflows.

The split keeps product data workloads separate from privileged infrastructure execution.

Security

  • Endpoint verification is read-only (GET/HEAD).
  • SSRF defenses reject local/private/link-local/documentation targets and unsafe URL schemes.
  • DNS results and redirect targets are revalidated.
  • Response size, redirect count, timeouts and concurrency are bounded.
  • Ingestion is authenticated in production.
  • Privileged VPS execution uses short-lived SSH certificates and host-key verification.
  • Secrets and private keys are not committed to this repository.

Commercial acquisition tool

npm run lead:audit -- https://example.com

The Lead-Leakage Audit performs a public-page, read-only conversion-path audit. It does not submit forms, authenticate, bypass controls or access private systems. It can be used as an evidence-first acquisition wedge for OSA services without becoming a separate product strategy.

Quick start

bash
npm installnpm testnpm start

Payments

The public checkout exposes one human product: the $79 OSA 30-Day MCP Reliability Pilot. A buyer can request a private invoice/bank-transfer route or pay immediately with USDC on a verifier-backed network. Card and mobile-wallet buttons stay hidden until a live merchant rail is independently verified. Tap is owner-disabled. Optional x402 v2 middleware remains the machine-payment path for /best and /score when OSA_PAY_TO is configured. Requests, test events, page views, and wallet connections are never counted as revenue.

Operating rule

OSA is one product. Agent Trust Oracle, Procurement Guard, pricing/benchmark intelligence, MCP reliability and payment-rail checks are modules of the same pre-transaction decision system.

Progress means verified external usage, verified payment and repeat usage — not internal test rows, crawler probes, generated dashboards or speculative revenue.

Revenue-first gate

Before first verified external payment, OSA keeps one primary revenue path plus one backup and freezes generalized product/infrastructure/payment/browser/autonomy expansion unless it is the minimum fix for a named external blocker or a real safety incident.

Every non-trivial change should carry the decision record defined in knowledge/05-revenue-gate.md. Pull requests are checked by the Revenue Gate workflow. The governing distinction is strict: BUILT != VERIFIED != SELLING != PAID.

Ecosystem

[Powered by RustChain]

Source: README.md at commit 609dbc7

Tools

1

procurement_decision

Use OSA before buying an API or MCP tool. Returns ALLOW, HOLD or DENY using budget policy, endpoint-scoped trust evidence, price history and optional cheaper equivalents. Get an OSA API key at https://pkctqxeydfuiupadaoov.supabase.co/functions/v1/procurement-site.

Parametersnameassetnetworkendpointmin_trustcapabilitymax_price_usdoffered_price_usd

Version history

1
  1. v0.2.2LatestSep 16, 2026

Score

90% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.