M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
A

Affiliate Networks MCP

Community
Ask your affiliate networks in plain English from your AI workspace.
How to install ↓ Compare GitHub source repository ↗
Category
Other #10 of 230
Stars
★ 7 Popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Node.js 20+ · Hosted (no local runtime)
Credentials
API key / credential required
License
MIT
Last commit
70FMRS · B

Community-driven open-source MCP server to integrate affiliate networks into AI workspaces. It offers broad adapter coverage but most are experimental, so caution is advised. Main strength is covering both publisher and advertiser sides, though advertiser side is read-only. A useful tool for marketers wanting to simplify affiliate data queries.

Strongest · Documentation 17/20 Weakest · Reliability 11/20

Reliability
11/20
Security and permissions
14/20
Maintenance
14/20
Documentation
17/20
Setup experience
14/20
Why each score
Reliability 11/20
README claims a vitest suite, CI smoke tests and `npm run verify`, but static review cannot verify that execution evidence; most of the 86 adapters self-declare as experimental, with wide behavioural variance and uncontrolled upstream API drift. Error handling is designed (NetworkErrorEnvelope, doctor diagnostics), but declared-tools-vs-real-behaviour cannot be confirmed from the supplied material. Scored within the static cap at 11; deductions: no verifiable execution evidence, heavy experimental surface, dependence on 72 third-party APIs.
Security and permissions 14/20
No red lines: credentials in a local 0600 file, telemetry opt-in and data-free, brand-side adapters enforced read-only (non-GET refused), no malware indicators. Deductions: README/manifest are untrusted evidence; hosted-connector vault and browser-driven dashboard session details (PRIVACY.md, security page) are not verifiable from supplied material; setup flows prompt users to paste GitHub tokens, widening the social-engineering surface; cache and vault guarantees lack source-level evidence.
Maintenance 14/20
MIT licence, version 0.20.0 with ongoing iteration, RELEASE/REPORT/CONTRIBUTING processes, 24h first-review commitment, and a security page. Deductions: 123 open issues against 6 stars suggests backlog or over-scoped scope; apparent single-maintainer ownership and the paid hosted tier's sustainability are unproven; no evidence of an independent security-disclosure channel.
Documentation 17/20
Documentation is exceptionally layered: install, dual-track onboarding, per-network setup pages, capability matrix, privacy contract, troubleshooting (doctor/test), client support-state table, and honest disclosure of limits (experimental, approval waits). Deductions: doc-vs-capability consistency is untrusted evidence; browser-driven session and hosted security details require external pages; many network docs are self-described 'built from public docs, untested'.
Setup experience 14/20
Three primary paths (.mcpb one-click extension, npx setup wizard, manual JSON config snippet), concrete commands for Claude Desktop/Code and Codex, with backup/uninstall instructions and honest labelling of untested paths. Within the static setup cap of 15, scored 14; deductions: Node 20 prerequisite, Cowork's private-mirror + org-admin complexity, other clients 'compatible in principle, untested', and manual .mcpb re-download updates.

Static review · not runListed 2026-09-08

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesUses the networkControls a browser

Best for

  • Affiliate publishers who want natural-language access to multiple networks
  • Agencies managing multiple brands' programs
  • Teams needing periodic reporting and anomaly monitoring

Not for

  • Users without an AI client (e.g., Claude Desktop)
  • Advertisers needing write operations (e.g., link creation) since advertiser adapters are read-only
  • Clients that are not yet officially supported (e.g., Cursor, VS Code) without testing
  • Networks where clicks are essential, as many adapters lack click support

Required permissions

  • Credentials for network APIs (API keys, tokens, etc.)
  • Optional: local file write to `~/.affiliate-mcp/.env` and `brands.json`
  • Optional: anonymous usage telemetry (off by default)
  • Optional: local result cache (off by default; writes files when enabled)

Risks and side effects

  • Experimental community adapters may be unreliable or return inaccurate data.
  • Credentials stored in local files; risk if file permissions are loose or device compromised.
  • Hosted connector stores credentials in encrypted vault but involves third-party service.
  • API calls may incur rate limits or costs depending on the network.
  • Untested client paths may not work; check support status table.

Setup

Before you start

Runtime：Node.js 20+ · Hosted (no local runtime)

Other optional settings (1)
AFFILIATE_MCP_CACHE optional Set to on in ~/.affiliate-mcp/.env to enable optional local result caching of closed reporting windows; off by default.

Technical track: install Node.js 20+, run npx affiliate-networks-mcp setup to configure credentials interactively, then npx affiliate-networks-mcp test to verify. Use npx affiliate-networks-mcp install to auto-configure Claude Desktop or Codex. For Claude Code, run claude plugin marketplace add bobberrisford/affiliatemcp and claude plugin install affiliate-networks-mcp@affiliatemcp. Non-technical users can download the .mcpb file and install in Claude Desktop. Also available via hosted connector at mcp.agenticaffiliate.ai (paid, OAuth).

Check that it works

After installing, ask "What affiliate networks do you have access to?" in a new Claude conversation — it should list every network you configured. If you registered brands, also ask "list my brands".

Troubleshooting

  1. If queries return nothing, check credentials are configured: run `npx affiliate-networks-mcp setup` or trigger the `affiliate-network-setup-help` skill.
  2. If stale npm cache, run `npx --yes affiliate-networks-mcp@latest` to fetch latest.
  3. If Cowork rejects the repository, ensure it's private or run `cowork-mirror` to create a private mirror.
  4. If a specific network fails, run `npx affiliate-networks-mcp doctor <network>` for diagnostics.
  5. Ensure API keys have no trailing spaces and tokens are not expired.

Things to try

Once connected, you can ask your AI assistant things like:

  • What did I earn across all networks last month?
  • Which programmes have transactions still pending after 90 days?
  • How is Acme's programme doing this quarter?
  • Audit the affiliate links in my sitemap at https://mysite.com/sitemap.xml

Use cases

Publisher asks 'What did I earn last month?' across all networks
Brand-side user asks 'How is Acme performing this quarter?'
Agency asks 'Show me revenue across all my clients this week.'
Ask 'Any anomalies in the affiliate data this week?'

Supported clients

Claude Desktop
Claude Code
Codex
Claude CoworkPartial support

Listed from the project's documentation, not tested by this site.

Overview

Affiliate Networks MCP is a local MCP server that lets you query multiple affiliate networks (both publisher and advertiser sides) using natural language. It supports over 80 network adapters, most community-built and experimental. Your credentials stay local in ~/.affiliate-mcp/.env. Includes skills for consolidated earnings, anomaly detection, and reports. Aimed at publishers, brands, and agencies who want to chat with their affiliate data.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision cba5aff01d09 Data synced 2026-10-19 Read the FMRS scoring method