Choose a remote MCP connection
In your AI client, choose Streamable HTTP or the equivalent remote MCP server option.
Model Context Protocol
The DoroSocial MCP server lets MCP-compatible agents such as Codex and Claude work directly with your approved DoroSocial brands, content, drafts, schedules, and publishing status without asking users to paste API tokens into chat.
Quickstart
In your AI client, choose Streamable HTTP or the equivalent remote MCP server option.
Enter https://doro.social/mcp. Leave bearer token and custom headers empty for the standard OAuth connection.
Click Authenticate. DoroSocial opens a secure sign-in and consent screen where you choose which brands the agent may access.
Approve the requested DoroSocial agent capabilities only if the brand selection and account are correct.
https://doro.social/mcpAdd this to your Codex config.toml, then run codex mcp login dorosocial and complete DoroSocial consent in your browser.
[mcp_servers.dorosocial]
url = "https://doro.social/mcp"Use this project .mcp.json configuration, or run claude mcp add --transport http dorosocial https://doro.social/mcp. Open /mcp in Claude Code and authenticate.
{
"mcpServers": {
"dorosocial": {
"type": "http",
"url": "https://doro.social/mcp"
}
}
}For a remote connector UI, use name DoroSocial, transport Streamable HTTP, and the server URL above. Leave token and custom headers empty for OAuth. No credential belongs in these examples; your client stores the credential after sign-in.
Server name: DoroSocial. Transport: Streamable HTTP (stateless JSON responses). You need a DoroSocial account with the API access entitlement, included in Pro by default. Account access policies can affect eligibility; connection consent checks current access. Compare plans.
MCP is the agent-friendly interface for DoroSocial. It exposes workflows as tools with useful names and structured inputs, so an agent can plan and manage social work without reverse-engineering raw API endpoints.
get_brand_context, update_brand_context, list_content_plans, create_content_plan, and update_content_plan help organize source work within permitted brands.import_media, begin_media_upload, complete_media_upload, and attach_post_media handle assets. get_content_transcript, request_transcription, estimate_generation_cost, and request_post_generation support source-based drafting.get_content_calendar, get_post, get_job_status, get_operation_status, get_post_performance, list_publish_errors, and list_audit_events expose current workflow state. summarize_recent_comments and save_comment_reply_draft support human-reviewed replies.Tool availability depends on granted scopes, brand access, feature eligibility, and whether writes are enabled. Ask for get_connector_capabilities before assuming an action is available.
list_brands and list_connected_accounts. Choose the intended brand and destination.search_content, read it with get_content_item, and prepare a platform draft with create_post_draft. Review the preview from get_post_preview.get_connector_capabilities and preflight_post, then requests review with request_post_approval in review mode.This illustrates the implemented tool flow, not a published customer result. AI generation and transcription consume account resources; connected-account eligibility, scopes, scheduling windows, and publishing limits still apply.
Most users only need the MCP server URL. MCP clients use OAuth discovery metadata behind the scenes to learn where to authenticate and how to request consent.
MCP endpoint
https://doro.social/mcpProtected resource metadata
https://doro.social/.well-known/oauth-protected-resource/mcpAuthorization server metadata
https://doro.social/.well-known/oauth-authorization-server/oauthMCP is best when a person connects Codex, Claude, or another agent and wants it to inspect context, draft content, and operate through consented DoroSocial tools.
The Public API is best for backend services, scripts, SDKs, and deterministic app-to-app integrations that should authenticate with scoped API tokens.
View API docsRead get_connector_capabilities for the selected brand’s current scheduling permission. In review mode, use request_post_approval; report the proposal as pending, include the returned Schedule link, and identify the brand, account, time, and timezone. Successful approval in Schedule completes the action without a follow-up agent call.
Only the owner can enable Allow automatic scheduling in Settings → Agent connections for a specific OAuth connection and brand. Chat consent does not change permission. Recheck capabilities before using schedule_post; enabling permission does not schedule existing proposals. Automatic mode covers future draft scheduling, not publish-now, retries, or changes to already-scheduled content. API-token MCP has no automatic-mode setting. REST integrations have their own authorization contract.
Clarify ambiguous brands, destination accounts, and times before writing. Never silently choose the first account or all channels. The MCP calendar excludes pending proposals; use known approval IDs with get_post_approval or review them in the browser. Calendar proposals remain pending beyond 24 hours; past times require a new future time.
Use returned states and next actions when explaining results. Pending is not scheduled, preflight is not approval, and queued is not published. Reread current state before treating an old operation receipt as current status. Tool guidance helps the agent communicate; server checks enforce permission.
Follow the approval and automatic-scheduling guideFor invalid input, inspect error.details.validation_errors. Each entry identifies a field and constraint, with applicable bounds, allowed values or format guidance. For example, sending 21 tags reports the 20-item maximum. Correct the listed fields; do not guess repeatedly. Submitted private values are not echoed.
Content creation accepts tag display names. The creation result and get_content_item return canonical tags with names and slugs. Search requires exact slugs: a tag named “Launch Campaign” is searched as launch-campaign. Multiple tag filters must all match. Tags are separate from social hashtags.
Content detail returns at most five asset references per page. Check asset_pagination for total_count, returned_count and page_size, and assets_complete for asset completion. Follow next_cursor until it is null to retrieve all description chunks, assets and permitted posts. Nine carousel images arrive as five then four; do not treat the first page as the complete carousel. If content changes during retrieval, restart without a cursor.
For a missing permission, inspect error.details.required_scopes. Ask the owner to reauthorize the connection with those scopes, or issue an appropriately scoped API token. Check get_connector_capabilities for supported permissions. Automatic scheduling does not grant missing scopes or bypass approval rules.
Revoke OAuth access using Disconnect in Settings → Agent connections (sign-in required). Revocation stops that connection; it does not delete drafts or scheduled posts already created. Review those separately.
Advanced integrations can also authenticate with a scoped DoroSocial API token in the HTTP Authorization header: Bearer YOUR_DOROSOCIAL_API_TOKEN. Replace the placeholder only in your client’s secure credential storage. Never paste it into a prompt or commit it. Revoke API tokens in Settings → API tokens. Token access follows its scopes and allowed brands; it does not offer the OAuth automatic-scheduling setting.
Leave it empty for the normal DoroSocial MCP flow. Authentication happens through the OAuth screen after you click Authenticate.
Confirm the URL is exactly https://doro.social/mcp, then retry authentication. If you are testing a local development server, use the matching local URL and make sure that server is running.
Reconnect and select the brand during consent, or review the existing connection from DoroSocial settings.