Theta MCP for AI. Manage AI Deployments and Video Workflows.

This server gives your agent direct command over decentralized infrastructure—everything from managing AI deployments on EdgeCloud to running full video pipelines and submitting blockchain transactions. You don't need a dedicated terminal; your agent handles it all.

AI Model Deployments:
Your agent can manage the entire lifecycle of AI models using tools like ai_list_standard_templates to see what built-in templates are ready for use, or ai_list_deployments to check every current and past deployment associated with your project. Need a new environment? You'll launch it by calling ai_create_deployment for a fresh AI model or Jupyter Notebook setup.

When you're done with the compute cycle, you can shut down resources using ai_stop_deployment to save cash, or delete the entire deployment via ai_delete_deployment. If a GPU Node gets stopped, you can bring it back online and ready for work by running ai_start_deployment; you'll also check available hardware types using ai_list_vm_resources.

Blockchain Operations:
Your agent lets you read the whole ledger. You can fetch detailed status and balances for any specific Theta account address with rpc_get_account. To track network movement, you can get finalized details for any block at a specific height using rpc_get_block_by_height, or pull complete data using a unique hash identifier via rpc_get_block.

For real-time monitoring, check the mempool contents to see all transactions waiting confirmation with rpc_get_pending_transactions, and drill down on specific events by getting transaction details with rpc_get_transaction using its unique hash. The network also gives you basic info like the current version number of the blockchain software running via rpc_get_version.

For sending money or data, your agent handles transactions securely. You'll list all local account addresses managed by the Theta CLI integration with cli_list_keys and create a new, encrypted key locally using cli_new_key. If you need to pause an account for safety, use cli_lock_key; when it’s clear everything is safe, run cli_unlock_key on that specific account.

You can send both Theta or TFuel tokens from an active local account with cli_send. For broadcasting raw transactions, you've got two options: submit a signed data payload immediately and wait for confirmation using rpc_broadcast_raw_transaction, or use rpc_broadcast_raw_transaction_async to send the data without forcing your agent to wait.

Key Management:
Beyond just sending tokens, you manage key security directly. You'll list all local account addresses with cli_list_keys; create a new encrypted account using cli_new_key; lock an existing account with cli_lock_key until it's safe to proceed; and unlock that account again when transactions are ready via cli_unlock_key.

Video Media Pipeline:
Your agent manages the entire media workflow. You can start by listing all videos tied to your service account using video_list, or search your entire library for specific assets based on metadata like date or tags, which you do with video_search. Before a big upload, generate a unique, temporary URL using video_create_upload_url; this is where external clients can securely drop large video files.

To process raw media, call video_transcode to convert files into multiple formats—say, 1080p MP4—for distribution. For live content, you'll first set up a new stream object using video_create_livestream, then assign an Edge Ingestor service using video_select_ingestor; remember to release that capacity later with video_unselect_ingestor. You can retrieve the metadata and status for any video asset or stream by running video_get, while checking available processing services with video_list_ingestors.