Adobe Launch MCP server: rules, data elements and libraries in Claude, Codex or Cursor
Adobe Launch reaches Claude, Codex or Cursor through the Agent Planners MCP endpoint with 7 reads and 4 writes of its own: properties, rules, data elements, extensions and libraries with their publishing state, and approval-gated rule switches and development libraries.
What already exists for Adobe Launch
Adobe Launch (now called Tags in Adobe Experience Platform) was not on Experience League's list of Adobe-hosted MCP servers in November 2026.
Under any layer sits the Reactor API (reactor.adobe.io, JSON:API): every change is a new revision, and nothing reaches a site until a library is built and published.
What this setup adds
| Calling the Reactor API yourself | Agent Planners MCP | |
|---|---|---|
| Launch tools | Whatever you write | 7 reads and 4 writes, plus two shared connection reads |
| Library adds after a change | 409 "needs to be revised" | Revised and added with the new revision |
| Builds, publishing, custom code, extensions | Sent when called | Held for a person on every key mode |
| Other platforms | Launch only | Google Tag Manager and every platform the workspace connected, on one endpoint |
How to configure it
- 1In Adobe Developer Console add the Experience Platform Launch API API to a project with an OAuth Server-to-Server credential, and put the credential on the product profile it should use.
- 2On Integrations open the Adobe Launch card, press Paste JSON (or Manual entry) and Test connection. The card requests only this product's scopes:
AdobeID,openid,read_organizations,additional_info.job_function,additional_info.projectedProductContext,additional_info.roles. - 3Create a key in API & MCP:
tools:readfor reading, addtools:writefor changes. To keep a key to Adobe, choose Only these platforms → Adobe. - 4Add the server to your client with one of the snippets below, run
describe_permissions, thenadobe__get_connectionandadobe__launch_list_rules.
# Claude Code
claude mcp add --transport http agent-planners https://www.agentplanners.com/api/mcp \
--header "Authorization: Bearer $AGENTPLANNERS_API_KEY"
# Codex (signs in with OAuth in the browser)
codex mcp add agent-planners --url https://www.agentplanners.com/api/mcp
# Cursor (~/.cursor/mcp.json)
{ "mcpServers": { "agent-planners": { "url": "https://www.agentplanners.com/api/mcp", "headers": { "Authorization": "Bearer ${env:AGENTPLANNERS_API_KEY}" } } } }The Adobe Launch tools
| Tool | Kind | What it does |
|---|---|---|
launch_list_companies | Read | The Tags (Reactor) companies the credential reaches (GET companies): id (CO…), name, org id. |
launch_list_properties | Read | Tags properties of the configured company (GET companies/{id}/properties): id (PR…), name, platform (web | mobile | edge), domains, enabled. |
launch_list_rules | Read | Rules of a property (GET properties/{id}/rules): id (RL…), name, enabled, published, dirty, revision. |
launch_list_data_elements | Read | Data elements of a property (GET properties/{id}/data_elements): id (DE…), name, delegate, enabled, published. |
launch_list_extensions | Read | Installed extensions of a property (GET properties/{id}/extensions): id (EX…), name, display name, version, enabled. |
launch_list_libraries | Read | Libraries of a property (GET properties/{id}/libraries): id (LB…), name, state (development | submitted | approved | rejected | published), build required, environment. |
launch_get | Read | ANY Reactor API GET (companies[/{id}], properties/{id}[/rules|data_elements|extensions|libraries|environments|hosts], rules/{id}[/rule_components], data_elements/{id}, libraries/{id}[/builds], builds/{id}, extension_p… |
launch_create_library | Write | Create a development library in a property (POST properties/{id}/libraries): { accountId?, propertyId?, name, ruleIds?, dataElementIds?, extensionIds? (the resources to add) }. |
launch_build_library | Write — always a person | Build a library (POST libraries/{id}/builds) — deploys it to the library's environment: { accountId?, libraryId }. |
launch_set_rule_enabled | Write | Enable or disable a rule (PATCH rules/{id} attributes.enabled — a new revision; it reaches the site only when a library with it is built and published): { accountId?, ruleId, enabled: true | false }. |
launch_mutate | Write | Any other Reactor write (method POST | PATCH | DELETE, path such as properties/{id}/rules, rules/{id}/rule_components, properties/{id}/data_elements, libraries/{id}/relationships/rules; body = JSON:API { data: { type,… |
describe_tool on one of them says to connect the Adobe Launch card. A key narrowed to the Adobe platform reaches only the Adobe products the workspace has connected; adobe__get_connection shows each one.What waits for a person, whatever the key allows
Ordinary writes — a development library, a rule switched on or off, settings that are not code — run inside the workspace's guardrails or queue for an admin, as the key was minted. These are held for a person on every key mode:
- A library build and submit / approve / publish.
- Custom code, and any rule-component or data-element settings change that could rewrite it.
- Extensions, environments and hosts.
- Any delete.
What we measured
- 13 properties read on a real Launch company, with rules, data elements and libraries.
- In a throwaway property: a data element created, renamed and deleted; a host; a development environment; a library with the Core extension; a build that succeeded. Nothing was published to a site.
Frequently asked questions
- Is there an official Adobe Launch MCP server?
- Not as of November 2026: Adobe Launch was not on Experience League's list of Adobe-hosted MCP servers. Agent Planners reaches the Reactor API with your own Server-to-Server credential.
- Can an MCP client publish Launch changes?
- It can ask; a build or a publish action is held for a person whatever the key's write mode.
- Is Adobe Launch the same as Adobe Launch?
- Yes — Adobe renamed Launch to Tags; the Developer Console API is still Experience Platform Launch API.