agentleFS
stevemarksavastevemarksava/superteam/skills/erp-connectors/SKILL.md

erp-connectors

Skill1 starsChanged 2 months agoMITView on GitHub

Connector selection matrix for agent integration with enterprise systems (ERP, CRM, HRIS, data platforms). Power Platform / Logic Apps / Custom MCP / Copilot Studio custom connector. ACP governance mandatory.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When designing an agent that integrates with any enterprise system, picking the wrong connector approach costs a sprint.

    line 11
  2. Distinctive:

    Check current MCP server availability: Before building a custom MCP server or custom connector, check whether the target system has a …

    line 39
  3. Distinctive:

    Build and deploy: Follow the appropriate downstream skill:

    line 58
  4. Distinctive:

    Route through Logic Apps when durability matters: If a connector operation can fail, needs retry logic, or involves multiple system calls, …

    line 75

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/erp-connectors — Connector Selection Matrix

When designing an agent that integrates with any enterprise system, picking the wrong connector approach costs a sprint. This skill provides the selection matrix and the governance check. Run it at blueprint stage — before Eric writes any integration code.

ACP MANDATORY: Advanced Connector Policies (ACP, GA June 4, 2026) replace classic Business/Non-Business DLP for certified connector governance in enterprise Copilot Studio agent builds. Architecture designed on the old DLP model alone is insufficient from June 2026. Verify ACP scope (certified connectors only — custom connectors still use classic DLP) at author time.

When to invoke

  • Designing an agent that integrates with any enterprise system: ERP (SAP, D365, BC, Oracle, NetSuite), CRM (Salesforce, Dynamics), HRIS, data platform (Snowflake, Databricks), or any internal API
  • Bob is blueprinting an agent architecture and needs a connector decision before the blueprint is locked
  • Nexus is producing a Platform Brief for a project with enterprise integration requirements
  • Client asks "how does the agent talk to [system]?" — this skill provides the structured answer and the governance check

Steps

  1. Enumerate target systems. List every external system the agent must interact with. For each, record:
    • System name and version
    • Available integration surfaces (REST API, OData, MCP server, Power Platform connector, SOAP/RFC for legacy)
    • Required operations (read-only vs read-write, which records or endpoints)
    • Auth mechanism (Entra ID, OAuth2, API key, SAML, XSUAA)
  2. Apply the connector selection matrix.

    Connector typeBest forAuth complexityCopilot Studio native?Notes
    Power Platform connector (1,000+ pre-built)M365, Dataverse, published SaaS integrations (Salesforce, ServiceNow, SAP, etc.)Low — Entra ID or OAuth2 managed by the connectorYes — native actionSimplest path for Copilot Studio agents; always check connector catalogue before building custom
    Logic Apps connector (400+ managed)Enterprise integrations requiring mid-layer processing: transformation, retries, long-running, durable executionLow-Medium — connector manages auth; Logic App orchestratesVia HTTP triggerUse when the chain involves multiple systems or needs retry/compensation logic; agent calls Logic Apps HTTP trigger, Logic Apps does the heavy lifting
    Custom MCP serverTarget system has no Power Platform connector, or agent needs a domain-specific tool surface (SAP S/4HANA, custom internal systems)High — vendor-specificVia MCP add (Copilot Studio creates a PP connector behind the scenes)MCP is the dominant emerging pattern — ERP vendors are publishing MCP servers. Check current vendor MCP availability before building a custom one
    Copilot Studio custom connectorTarget has a REST API but no MCP wrapper and no certified PP connectorMedium — OpenAPI spec, OAuth2 or API keyYes — configure from OpenAPI specRight bridge when the target is REST-capable but not MCP-ready; simpler than building a full MCP server
  3. Check current MCP server availability. Before building a custom MCP server or custom connector, check whether the target system has a vendor-published one. Verify this table at author time — the MCP ecosystem is evolving monthly.

    PlatformMCP status (September 2026)Auth
    Business CentralGA 2026Entra ID
    SnowflakeGA — managed in-accountSnowflake OAuth + RBAC
    DatabricksPublic Preview (January 2026)Entra ID + Unity Catalog
    TeradataGATeradata auth
    SAP Commerce CloudQ2 2026 target — verify current availabilitySAP IAS + XSUAA
    Work IQ (M365 semantic index)PreviewEntra ID
    Salesforce / MuleSoftNot published as of September 2026—
    Oracle NetSuiteNot published as of September 2026—
  4. Verify ACP governance. For every connector approach:
    • ACP (Advanced Connector Policies, GA June 4, 2026) applies to certified Power Platform connectors — default-deny allowlist, action-level control within a connector, can block specific MCP servers
    • Custom connectors (OpenAPI-based) still fall under classic DLP — ACP does not yet cover custom connectors; verify this boundary at author time
    • MCP servers added to Copilot Studio create a Power Platform connector behind the scenes — they inherit ACP/DLP governance; specific MCP servers can be blocked by ACP. This is not obvious from the Copilot Studio user-facing flow.
    • Purview sensitivity labels propagate through connector data paths for regulated clients — factor this into governance design for any data classification requirement
  5. Build and deploy. Follow the appropriate downstream skill:
    • BC MCP integration: skills/bc-mcp/SKILL.md
    • SAP BTP integration: skills/sap-btp-agents/SKILL.md
    • Copilot Studio agent authoring: skills/copilot-studio-build/SKILL.md
    • Azure AI agent: skills/azure-ai-foundry-dev/SKILL.md

Deliverable format

  • Connector selection decision document citing this matrix and the project-specific reasoning for each integration point
  • ACP policy record: which connectors and which actions are allowed per environment (dev/test/prod)
  • For each integration point: connector type chosen, auth mechanism, read/write scope, licensing note

Non-negotiables

  • ACP is mandatory, not optional. Any enterprise Copilot Studio + connector build that governs using only the old Business/Non-Business DLP model is architecturally insufficient as of June 2026. Design ACP policies at blueprint stage.
  • Check MCP server availability before building custom. The default assumption is that a vendor-published MCP server may exist. Building a custom MCP server when a vendor one is available wastes a sprint.
  • Verify ACP scope at author time. ACP currently covers certified connectors only — custom connectors still use classic DLP. This boundary may shift; verify against current Power Platform documentation at the start of every engagement.
  • Route through Logic Apps when durability matters. If a connector operation can fail, needs retry logic, or involves multiple system calls, route through Logic Apps — not directly from the agent. Agents are not durable executors.
  • Brief G on connector licensing before scaling. Power Platform connector licensing (per-user, per-flow) and Logic Apps (Consumption vs Standard) have significant cost differences at scale. G must model costs before any connector architecture goes to production.

References

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.