MCP Studio

Connect your systems.Give AI useful tools.

Create, test, and version MCP for AI access to enterprise systems, with workspace caller scopes and an assessed delivery path to AWS.

Request an MCP Studio demo Explore capabilities

MCP
REST / OpenAPI
AWS Lambda
AgentCore Gateway

Which data can AI use, through which tool?

MCP lets apps and AI agents use tools, resources and prompts from connected systems. MCP Studio helps your team author, test and version those capabilities before activation.

AI AgentRequest product DEMO-001

MCP Toolget_product(sku)Authored and managed in MCP Studio

Source systemAn API configured for connection

MCP · System connections

For teams connecting AI to business systems

AI development, integration and cloud teams can prepare tools, inspect drafts and select releases in one workspace, while assessing access and deployment needs.

Capabilities

Create, test, version, and scope MCP access in a workspace for technical teams.

Shape a tool. Inspect its input and output.

Use the guided builder or advanced code for Tools, Resources, and Prompts. Test tools through forms or JSON before building a release.

Studio / get_product

Tool definitionDraft

tool get_product {
  input: { sku: string }
  output: {
    name: string,
    stock: number
  }
}

Example input → outputDEMO-001 stock: 12

Test before Build

Studio · Draft & Test

Get help while developing a draft

Open Draft with AI in Studio for assistance powered by AWS Bedrock. Optionally attach saved draft source as context.

Review suggestions and test before use; automatic creation, saving and testing across the full workflow are not verified.

Studio / AI assistant

AWS BedrockSource: opt-in

get_product({
  sku: "DEMO-001"
})get_product({
  sku: "DEMO-001"
})get_product({
  sku: "DEMO-001"
})

Check found and not-found cases before use.

Studio · AI assistant

Turn an API into a tool

Configure a connection, import OpenAPI JSON/YAML or define REST operations, map input and response fields, then test before adding the tool to a draft.

SheetsExperimental · Workspace

PostgreSQLExperimental · Workspace

REST / OpenAPI/products/{sku}

MCP Toolget_product(sku)

Input: sku
Output: name · stock

Select fields → test → add to Draft

Integrations

Separate drafts and active versions

Build and check a draft, create an immutable release with digest and revision references, then activate it or select an earlier release.

Rollback does not undo changes already made in external systems

Release AWorkspace

Draft
Build
Release
Activate

Release A activeContent sealed from Build

Draft remains editableRelease B available for later activation

Rollback does not undo external data changes

Releases & delivery

Test as a caller

Use Playground to test the active workspace release with a selected caller identity, separately from draft tests and AWS endpoint tests.

Support AgentWorkspace

tools/call

get_productskuDEMO-001

Illustrative test request

response

{
  "name": "Demo A",
  "stock": 12
}

Illustrative output

get_product In scope · update_product Out of scope

Playground

Prepare delivery to AWS

Verify STS, prepare a package and CDK template, review and deploy MCP as Lambda, with a path to aggregate targets through Amazon Bedrock AgentCore Gateway.

Management installs separately; AWS Managed REST accepts approved public HTTPS, not private networks

Build → Release

WorkspaceApplication hostConsumer scopes

AWSLambda → AgentCoreIAM / SigV4

AWS Managed REST: public HTTPS only
Workspace has no sandbox · token revocation does not revoke AWS access

AWS delivery

Choose what each agent can call

Set workspace consumer scopes at MCP/tool level, with credential expiry and revocation, separately from AWS IAM access.

Per-consumer UTC minute/day quotas; UI specifies 429 / Retry-After when full

ConsumerSupport Agent

get_productIn scope

update_productOut of scope

Per-consumer quotaUTC minute / day429 · Retry-After

Gateway & access

Activity & monitorTrace calls and outcomes

Review retained workspace request metadata and management audit to investigate call context and outcomes. This history excludes secrets and raw results.

Activity retains up to 1,000 recent events, not lifetime history or CloudWatch metrics.

Activity / Workspace

Request historySupport Agent

!
get_productRequest errorRequest metadata
update_productDeniedReview consumer scope

1,000recent events maximum

Not actual logs · no raw arguments/results shown

Activity & monitor

Start with the data to connect

REST/OpenAPI provides a Connection → Mapping → Test → Add tool flow. The following connectors have specific limits.

Google Sheets

ExperimentalWorkspace only

Fixed-range reads: up to 200 rows, 26 columns, and 2,000 cells; Google sign-in requires OAuth app setup

Microsoft Excel

ExperimentalWorkspace only

Business .xlsx on OneDrive/SharePoint: up to 200 rows, 26 columns, and 2,000 cells; no personal OneDrive or app-only auth

PostgreSQL

ExperimentalWorkspace only

Read-only access to approved tables/columns: up to 100 rows and 20 columns; no views or automatic password rotation

Experimental · Workspace only

These three connectors do not support Lambda or Gateway delivery. Row, column, and cell caps apply separately; do not multiply them. Validate against your data and permissions before use.

Choose one pilot use case

Evaluation examples, not enterprise customer results.

Integrations · REST / OpenAPIHow an agent requests API data

AI Agent

Find one example article

list_posts
REST API

Illustrative data for the agent

{ "id": 1, "title": "Demo post" }

Integration · Demo example

Expose a REST API to an agent

Use Connection, Mapping, and Test with synthetic data such as the JSONPlaceholder connection observed in the system. The report did not verify a live call.

Integrations · Google SheetsRead products from a cell range

Google SheetsExperimental

Read only · range A1:B3
SKU	Product name
DEMO-001	Example product A
DEMO-002	Example product B

Workspace only

Operations · Hypothetical example

Read a product list from a sheet

Configure a Google Sheets tab and range, then build a tool for an agent to try within the workspace. This connector is experimental.

AWS delivery · Release reviewWhat to review before AWS delivery

1
Confirm the releaseBuild → Release
2
Review target and permissionsAWS · IAM
3
Test the actual endpointEndpoint test

Checks still required, not a passed test

Development and cloud · Hypothetical

Review a release before AWS delivery

Separate Draft, Build, and Release, choose the delivery target, and test the endpoint to verify the version and execution path.

Prepare a demo around your systems

Installation time, delivery ownership, and training scope need a joint assessment.

Getting started · PreparationPrepare three things first
Source data
API/spec or data source
Owners
Trusted MCP authors and administrators
Callers and access
Consumer and permitted tools
Choose data and owners
Define a use case, API/spec or data source, trusted authors, administrators, and caller scope.
Use case · API · Authentication
Getting started · Build & ReleaseFrom connection to release
1. 1
  ConnectionSelect input / output
2. 2
  DraftTest the draft
3. 3
  Build → ReleaseCreate a sealed release
Build and test a draft
Configure a connection, map and test a tool, then build and create a release.
Connect · Draft · Test · Build
Getting started · Runtime testChoose a path, then test that path
Choose one path
WorkspaceTest with a Workspace consumer
AWSTest the AWS endpoint
Verify actual calls
Choose a path and verify
Assess workspace or AWS delivery and permissions. Test actual calls before agreeing PoC scope.
Release · Workspace / AWS · Verify

Technical information

Support and limitations
Intended audience	AI development, integration, and cloud teams with trusted authors and administrators
Runtime	Workspace on application host; Lambda / AgentCore are separate AWS paths
Protocol	Workspace UI displays 2026-07-28 · Streamable HTTP; no conformance/client test evidence
Workspace access	Consumer scopes · Bearer / machine OAuth · UTC minute/day quotas; UI specifies 429 / Retry-After when full
Cloud access	Gateway-wide IAM is separate from Workspace; no Cognito/Cedar or per-user cloud rights. Workspace token revocation does not revoke cloud access
Code boundary	Trusted authors only; no sandbox or multi-tenant isolation
Network	AWS Managed REST: approved public HTTPS only; no private networks, local fixtures, or HTTP
Activity	Up to 1,000 recent events, not lifetime usage or CloudWatch metrics
Local trial	Windows trial, machine-only with no tunnel/proxy sharing; no accepted cross-version data upgrade

More detail for your IT team

Based on UI and inline documentation reviewed on 20 Oct 2026. The review opened and read pages; it did not exercise MCP, connectors, AI clients, Bedrock, or AWS.

Frequently asked questions

What to know before choosing a setup and requesting a demo.

Try MCP Studio free for 30 days

Choose one real team task and assess the journey from authoring a tool to activation.

30daysFree trial

Explore Studio, Playground and Activity & monitor
Prepare MCP and data for your pilot
Confirm trial terms with our team before starting

Request a trial View plans

Contact our team to confirm trial terms, including AWS and external service costs, before starting.

A suggested 30-day evaluation

Day 1
Create your first MCP in Studio
Prepare a tool and draft around a team pilot.
Week 1
Inspect inputs and results
Test the draft and separately inspect an active release in Playground.
Weeks 2–3
Choose a release and caller scope
Review a release, activate it in Workspace and inspect Activity.
Week 4
Assess AWS delivery and next steps
Review permissions, AWS costs and PoC criteria before delivery.

Adapt the plan with our team to suit your systems and data.

Choose the right approach for your team

Discuss licensing, implementation and support scope before agreeing a service.

Trial

Free30 days

Start with a pilot before choosing a longer-term approach.

Studio · Playground
Workspace runtime
Activity & monitor

Request a trial

Team

Discuss shared MCP usage for your team.

Build · Release · Activate
Consumers · Gateway & access
Assess access and service scope

Enterprise

Assess deployment and integration for your organization.

AWS Lambda · AgentCore Gateway
Review network and AWS IAM
Define scope with eCloudvalley

Usage, implementation and support scope will be defined in your proposal.

Start with the MCPyour use case needs.

Bring your use case, source systems, AI client, and access requirements to assess fit and define a demo scope.

Request an MCP Studio demo

More to explore

AI Lens for Claude Enterprise AI Lens for Bedrock ECV Enterprise Assistance

Connect your systems.Give AI useful tools.

Which data can AI use, through which tool?

Capabilities

Shape a tool. Inspect its input and output.

Get help while developing a draft

Turn an API into a tool

Separate drafts and active versions

Test as a caller

Prepare delivery to AWS

Choose what each agent can call

Activity & monitorTrace calls and outcomes

Start with the data to connect

Google Sheets

Microsoft Excel

PostgreSQL

Choose one pilot use case

Expose a REST API to an agent

Read a product list from a sheet

Review a release before AWS delivery

Prepare a demo around your systems

Choose data and owners

Build and test a draft

Choose a path and verify

Technical information

More detail for your IT team

Architecture and data flow

What the AI assistant receives

Installation and private networks

Protocol and AI clients

Connector scope

Quotas and performance

Errors and continuity

Access and credentials

Operations and Local trial

Builds, releases, and documentation

Frequently asked questions

Do we need a technical team?

Do we need to host MCP ourselves?

Can every connector run on AWS?

Can we connect private-network systems?

Has production readiness been tested?

What are pricing, SLA, and installation timelines?

Try MCP Studio free for 30 days

A suggested 30-day evaluation

Create your first MCP in Studio

Inspect inputs and results

Choose a release and caller scope

Assess AWS delivery and next steps

Choose the right approach for your team

Trial

Team

Enterprise

Start with the MCPyour use case needs.

More to explore