MCP Server

WSO2 Agent Manager (AMP) ships a Model Context Protocol (MCP) server that lets MCP-capable AI assistants — Claude Code, Codex, Cursor, and others — drive the platform directly from the developer's terminal. Through the server, an assistant can list projects, register agents, trigger builds, and deploy images without leaving the conversation.

Overview

The MCP server is exposed by agent-manager-service at the /mcp path on the public API URL. It speaks MCP over Streamable HTTP and is protected by the same OAuth 2.0 Authorization Code + PKCE flow that secures the rest of the AMP API, against the bundled Thunder identity provider.

A single OAuth client (am-mcp) is pre-registered in Thunder out of the box, so most users do not need to provision their own client to get started.

Each tool is authorized against the same amp:* scopes as its REST API equivalent — for example, build_agent requires amp:agent:build. See Authorization for the full scope catalog, the predefined roles, and how the model applies across the Console, REST API, and MCP servers.

The caller's organization is always derived from the OAuth token — no tool takes an organization name or handle as input, and every operation is scoped to the organization the caller authenticated against.

Available Tools

The MCP server provides 13 tools organized into five toolsets covering the agent lifecycle — from creating a project through deploying an agent. Trace, log, and metrics tools live on a separate server; see Observer MCP Server.

Project Toolset (3 tools)

• list_projects — List projects in the caller's organization. Supports pagination via limit and offset.
• create_project — Create a new project (the logical container for agents and related resources).
• list_project_agent_pairs — Walk every project in the organization and return (project, agent) name pairs, with optional case-insensitive substring filters on project and agent names. Useful when an assistant needs to locate an agent without knowing which project it lives in.

Agent Toolset (3 tools)

• list_agents — List agents in a project. Each entry includes the provisioning type (internal or external) so the assistant can tell platform-hosted agents apart from externally-hosted ones.
• create_external_agent — Register an externally-hosted agent. Returns the agent identity, a long-lived API token, and language-specific zero-code instrumentation instructions (Python or Ballerina) the assistant can hand back to the user. The token is scoped to an environment: pass one via the optional environment parameter, or omit it when the organization has a single environment (with several environments configured, the parameter is required — use list_environments to discover valid names).
• create_internal_agent_python — Create a platform-hosted Python agent from a GitHub repository. Accepts repo URL, branch, app path, Python version, run command, interface type (DEFAULT chat-on-/chat:8000 or CUSTOM with port + OpenAPI), environment variables, and an optional auto-instrumentation toggle. Creating the agent automatically triggers its initial build.

Build Toolset (3 tools)

Internal agents only — external agents are never built by the platform.

• list_builds — Paginated list of builds for an agent, including status, image ID, and timestamps. Includes a retry_after_seconds hint when a build is still in progress.
• get_build_details — Detailed view of a single build: steps, durations, commit, and build parameters.
• build_agent — Trigger a fresh build from a specific commit (defaults to the latest commit on the configured branch). Returns immediately — successful builds trigger deployment automatically, so the assistant only needs to poll get_build_details for completion.

Deployment Toolset (3 tools)

• list_deployments — An agent's deployments across all environments, keyed by environment name, with current state (active, in-progress, failed, not-deployed, or suspended) and image.
• deploy_agent — Deploy a built image to the lowest environment in the deployment pipeline. Accepts runtime environment variables (plain values, sensitive flags, or references to existing secrets) and an enable_auto_instrumentation toggle.
• update_deployment_state — Transition a deployment in a specific environment: redeploy (fresh rollout of the current deployment) or undeploy (remove the deployment from that environment).

Environment Toolset (1 tool)

• list_environments — Paginated list of the organization's environments, with name, display name, and a production flag. Use it to discover valid environment names for tools that take an environment parameter, such as update_deployment_state and create_external_agent.

Finding the MCP Server URL

The MCP server URL follows the public URL of the agent-manager-service:

<server-public-url>/mcp

For a Quick Start or k3d install via the bundled Helm chart, the API is served through the OpenChoreo control-plane gateway, so the MCP server URL is:

http://api.amp.localhost:8080/mcp

The docker-compose dev stack publishes agent-manager-service straight onto a host port instead (SERVER_PUBLIC_URL=http://localhost:9000), so there the URL is:

http://localhost:9000/mcp

Each install registers only the origins it actually serves: make setup registers both (the gateway origin and the dev host port), while a Quick Start or k3d install registers the gateway origin alone. If you have changed serverPublicURL (Helm) or SERVER_PUBLIC_URL (docker-compose), substitute that value — and see the warning below.

Moving the public URL is a two-place change

MCP clients send the server's public URL — with a trailing slash — as the OAuth resource parameter (RFC 8707), and Thunder both validates that value against its registered resource servers and stamps it into the token's aud. Changing the public URL therefore means changing two settings together, in two charts:

Setting	Chart	Value
agentManagerService.config.serverPublicURL	wso2-agent-manager	the public base URL
thunder.bootstrap.agentManagerMcpBaseUrl	wso2-amp-thunder-extension	the same base URL

On docker-compose there is no chart to derive anything, so all three move together by hand in deployments/docker-compose.yml: SERVER_PUBLIC_URL, the matching KEY_MANAGER_AUDIENCE entry (the same URL plus a trailing slash), and the identifier registered in Thunder.

Leaving agentManagerMcpBaseUrl behind makes the browser redirect fail with invalid_target: The resource parameter does not match any registered resource server.

The matching aud entry needs no separate setting: wso2-agent-manager appends serverPublicURL plus a trailing slash to keyManager.audience when it renders, so that list only carries client IDs. (Charts released before that derivation list the URL literally — there, moving the public URL is a three-place change.)

Identifiers are matched exactly, so a deployment reachable on more than one origin needs one resource server per origin. thunder.bootstrap.agentManagerMcpDevBaseUrl is the opt-in second origin for the docker-compose dev stack; make setup-openchoreo sets it, and other installs leave it empty.

Default OAuth Client

AMP ships a pre-registered OAuth application in Thunder for MCP clients:

Client ID	Grant Type	Redirect URIs	Use Case
am-mcp	Authorization Code + PKCE	http://127.0.0.1:33418/callback, http://localhost:33418/callback	Interactive use with AI assistants that support browser-based OAuth (Claude Code, etc.)

The client is public (no secret), PKCE is required, and the callback port is pinned to 33418. Because the redirect URIs are baked into the OAuth client registration, every MCP client must run its local OAuth listener on port 33418 to complete the login.

The am-mcp client is provisioned automatically by the wso2-amp-thunder-extension Helm chart (script 59-am-mcp-client.sh) and is already wired into KEY_MANAGER_AUDIENCE on the docker-compose setup, so no manual registration is required for the default installations.

Upgrading an existing install

Both the MCP resource-server identifiers and the permissions that become MCP token scopes are seeded by 60-amp-resource-server.sh, which lives in a ConfigMap annotated helm.sh/hook: pre-install. Helm does not re-apply that ConfigMap on helm upgrade, so upgrading the chart leaves the old scripts in place and re-running the setup job would just re-seed the old values.

To pick up either, reinstall the amp-thunder-extension release (or replace the amp-thunder-bootstrap ConfigMap, then re-run the job). Symptoms if you skip it: an install seeded before the gateway origin was added registered only http://localhost:9000/, so a Quick Start or k3d client fails authorize with invalid_target; an install seeded without the permissions issues MCP tokens with an empty scope list and every tool call returns 403. Identifiers no longer in use are inert and can be left in place.

Configuring AI Assistants

The examples below assume the default local URL http://api.amp.localhost:8080/mcp. Replace it with your deployment's value if different.

Claude Code

Register the MCP server with Claude Code using the am-mcp client and the pinned callback port:

claude mcp add --transport http agent-manager http://api.amp.localhost:8080/mcp \
  --client-id am-mcp \
  --callback-port 33418

• --transport http selects MCP Streamable HTTP transport.
• --client-id am-mcp matches the pre-registered Thunder OAuth client.
• --callback-port 33418 pins Claude Code's local OAuth listener to the port registered as a redirect URI in Thunder.

Verify the registration from within a Claude Code session:

/mcp

The first tool call in any session opens a browser to the Thunder login page. Subsequent calls reuse the cached token until it expires; Claude Code refreshes it transparently.

Codex

Codex CLI consumes MCP servers through its ~/.codex/config.toml file. Because Codex's native MCP transport does not currently drive a browser-based OAuth flow, use mcp-remote to bridge Codex's stdio transport to the AMP server's HTTP endpoint — it will open the browser for OAuth on the first call and cache the token.

Add this block to ~/.codex/config.toml:

[mcp_servers.agent-manager]
command = "npx"
args = [
  "-y",
  "mcp-remote@latest",
  "http://api.amp.localhost:8080/mcp",
  "--client-id", "am-mcp",
  "--callback-port", "33418",
]

Restart Codex (or start a new session). On the first tool call, mcp-remote opens the Thunder login page; after authenticating, the token is cached under ~/.mcp-auth/ and reused until it expires.

tip

mcp-remote requires Node.js (and npx) on PATH. If you'd rather install the bridge globally instead of running it via npx -y, run npm install -g mcp-remote and replace command = "npx" / args = ["-y", "mcp-remote@latest", ...] with command = "mcp-remote" / args = [...].

Next Steps

• Try a few prompts in Claude Code, e.g. "List the projects in my org", "Create a new Python agent from this repo", "Deploy the latest build of agent X".
• See Observer MCP Server to let the same assistant inspect runtime logs, metrics, and traces for a deployed agent.
• See Observe Your First Agent for an end-to-end walkthrough of registering and instrumenting an agent — the same flow the MCP create_external_agent tool automates.