Workflow guides

Architecture comparison

MCP or an API: which connection does your project need?

An API lets software exchange information or request actions. MCP gives compatible AI assistants a standard way to discover and use selected tools and information. MCP often sits on top of an API. Choose the simplest connection that serves the task, while keeping authentication, permissions, validation, and work records.

Good fit
Choose MCP when multiple compatible AI hosts need to discover and use a bounded capability with descriptions and structured context, while the organization keeps an independent source service underneath.
Pause when
Choose a direct API or internal function when the consumer is deterministic application code, latency and throughput dominate, or exposing model-oriented discovery adds no useful interoperability.

When this approach helps

  • The capability should be usable from more than one assistant or agent host.
  • The consumer benefits from resource descriptions and tool schemas at discovery time.
  • The business service already has or needs an interface independent of any model host.

Follow these steps

  1. 01

    Identify what will connect

    List browsers, background jobs, other systems, AI assistants, and people separately. The same task may need an API for applications and an MCP connection for assistants.

  2. 02

    Keep the business rules in one place

    Keep validation, permissions, duplicate-request handling, and state changes in the shared service. The MCP adapter should translate requests without creating a second copy of those rules.

  3. 03

    Compare practical requirements

    Check discovery, compatibility, streaming, batch sizes, speed, versions, caching, rate limits, retries, customer separation, and logging. Choose the simplest connection that meets the requirements.

  4. 04

    Test the service and the AI connection

    Test the underlying service first, then use the MCP tool through a compatible assistant. Check structured errors and confirm that the adapter cannot grant extra permissions.

Documents and records to keep

Consumer and interface map

A diagram ties each consumer to the API, MCP adapter, data source, identity boundary, and owning team, revealing duplicated logic or accidental coupling.

Capability contract

The contract defines business inputs, outputs, errors, side effects, idempotency, service limits, and audit fields independently of the transport used by a client.

Compatibility test matrix

Tests cover direct service calls and supported MCP hosts, including discovery, schemas, authorization failures, partial outages, timeouts, and representative successful cases.

Common questions

Does MCP replace REST or GraphQL APIs?

Usually not. MCP gives AI hosts a standard consumption interface, while REST, GraphQL, RPC, queues, or database access may still power the owned service beneath it. The best layering avoids duplicating business rules in the adapter.

Can the same capability support both MCP and an API?

Yes. Put the capability and policy in a shared service, then expose purpose-built adapters. The API can serve deterministic clients while MCP supplies model-oriented discovery and structured invocation for compatible hosts.

Is MCP only useful for tool calls?

No. MCP can expose resources for context and prompts for reusable interaction patterns as well as tools for actions. Prefer the least powerful primitive that completes the consumer’s job.

AI workflow system

Find the task where AI could help.

Start with one repeated task. Agree on what AI can do, who checks its work, and how to measure the result.

Verified field result

Automation prepared the evidence. Human judgment still decided.

The system gathered evidence for reviewers. A person still decided whether to approve or reject each submission.

49/ 50
49 of 50 selected cases produced usable evidence packets for human decision.
Workflow
Marketplace template review
Receipt
#FR-2026-01
Verified
May–June 2026
Changes to external systems
0
  1. Map the frictionList the tools, steps, and person responsible.
  2. Test one pathTest the routine work and check that approvals still work.
  3. Keep the proofReview what ran, what stopped, and what needs recovery.