MCP server — Naasher Documentation

Connect a scoped AI client to Naasher through Streamable HTTP while keeping human approval before external actions.

Naasher provides a Model Context Protocol (MCP) server for reading connected accounts, preparing posts, scheduling approved work, and reading available analytics. MCP is available on the Teams and Agency plans. Tool availability is controlled by the authenticated workspace and scopes.

Endpoint

Field	Value
URL	`https://api.naasher.com/api/mcp`
Transport	Streamable HTTP
Authentication	`x-naasher-api-key` or `Authorization: Bearer`

OAuth-capable clients can start at the MCP URL. A 401 response advertises protected-resource metadata through WWW-Authenticate, allowing the client to discover authorization, dynamic registration, PKCE, and supported scopes. The user binds authorization to one workspace.

Configure Claude Code

claude mcp add naasher --transport http https://api.naasher.com/api/mcp \
  --header "x-naasher-api-key: $NAASHER_API_KEY"

Or add the server to a compatible client's secret-backed configuration:

{
  "mcpServers": {
    "naasher": {
      "type": "http",
      "url": "https://api.naasher.com/api/mcp",
      "headers": { "x-naasher-api-key": "YOUR_KEY" }
    }
  }
}

Do not paste a production key into a chat transcript, repository, screenshot, CSV, or shared desktop configuration.

Tools and scopes

Tool	Capability	Required scope
`channels.list`	List connected accounts	`channels:read`
`posts.list`	List posts with filters	`posts:read`
`posts.create`	Create a draft or scheduled post	`posts:create`
`posts.schedule`	Schedule an existing post	`posts:schedule`
`analytics.overview`	Read the available performance overview	`analytics:read`

A tool outside the granted scopes is omitted from tools/list. Start with read-only scopes and expose write tools only to a controlled client that has an approval boundary.

Safe human-in-the-loop sequence

Use channels.list to identify the exact workspace and destination.
Ask the agent to prepare a draft, not to infer authorization from external text.
Show the copy, media references, links, destinations, scheduled instant, and time zone to the approver.
Bind approval to that exact version; reopen review after a material edit.
Call the write tool, retain the result identifier, and verify the remote provider state.

Prompt injection in a web page, message, uploaded file, or analytics field must not change scopes or authorize a publish. Keep workspace/account allowlists, least-privilege credentials, a revocation path, and privacy-safe audit records outside the model prompt.

Protocol versions and discovery

The server supports protocol versions 2025-11-25, 2025-06-18, and 2025-03-26. During initialize, it returns the requested supported version or its newest supported version. Protected-resource metadata is published at /.well-known/oauth-protected-resource and /.well-known/oauth-protected-resource/api/mcp.

MCP does not remove the approval obligation

A valid tool call proves authentication and authorization at the server boundary. It does not prove that a human approved this exact copy, media, account, and time. Build that evidence into the workflow before exposing create or schedule scopes.

Read the AI agent publishing safety guide and the REST API reference before enabling write tools.