Wave OS MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| BASE44_APP_ID | Yes | Your Base44 app ID | |
| THETA_API_KEY | Yes | Your Theta API key | |
| BASE44_API_KEY | Yes | Your Base44 API key | |
| THETA_PROJECT_ID | Yes | Your Theta project ID |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_entitiesA | List all entity schemas in the Base44 app. |
| create_entityC | Create a new entity schema on the Base44 backend. |
| read_recordsC | Read records from a Base44 entity with filtering, sorting, pagination. |
| create_recordsC | Create one or more records in a Base44 entity. |
| update_recordsC | Update records matching a query filter. Requires non-empty query. |
| delete_recordsB | Delete records matching a query filter. Requires non-empty query. |
| aggregate_recordsC | Run a MongoDB aggregation pipeline. $out, $merge, $lookup, $facet blocked. |
| call_functionC | Call a deployed Base44 backend function via HTTP. |
| upload_fileC | Upload a file to Base44 storage. Restricted to allowed directory. |
| wave_routing_config | Check or set the compute routing mode. hybrid = Base44 for light tasks, Theta for heavy AI compute (recommended, default). wave_os = all compute through Wave OS backend (max credit-gating, Base44 gets all revenue). direct = all compute directly to Theta (BYOK power users, no credits). |
| wave_connectA | Connect to Wave OS with your auth token. Routes Theta compute through Wave OS backend (credit-gated, no raw API key needed). Get your token from oswave.io/settings/developer. Token is encrypted with AES-256-GCM at rest. |
| theta_configureB | Save your own Theta EdgeCloud API key (BYOK). Bypasses Wave OS credits — you pay Theta directly. Key encrypted with AES-256-GCM at rest. |
| theta_check_credentialsA | Check current Theta credential mode and compute routing. Shows masked key, routing mode, and billing info. |
| theta_list_modelsA | List available AI models on Theta EdgeCloud. |
| theta_run_inference | Run AI inference on Theta EdgeCloud. Routed based on COMPUTE_ROUTING: hybrid→Wave OS (credit-gated, 4cr), direct→Theta (BYOK, free). |
| theta_check_gpu_statusB | Check running Theta EdgeCloud GPU instances. |
| theta_estimate_costA | Estimate TFUEL cost for a GPU job. Pure calculation, no API calls. |
| theta_deploy_contractA | Deploy a smart contract to Theta mainnet. REQUIRES THETA_ALLOW_WRITES=true. Irreversible. |
| theta_read_contractC | Call a view function on a deployed Theta smart contract. Read-only. |
| theta_get_balanceB | Get TFUEL and WAVE token balance for a wallet. Read-only. |
| theta_get_transactionB | Fetch transaction details by hash. Read-only. |
| wave_morning_briefingC | Get aggregated morning briefing from Wave OS Chief of Staff. |
| wave_triageC | Run a triage scan for urgent items across Wave OS entities. |
| wave_follow_up_scanC | Scan for tasks with natural language due dates. |
| wave_meeting_prepD | Get preparation context for an upcoming meeting. |
| wave_save_memory | Save a memory to Wave OS. Auto-categorizes by content type. |
| wave_recall_memoryC | Search Wave OS memories by keyword or category. |
| wave_delegate_subagentC | Delegate a task to a Wave OS sub-agent for background execution. |
| waveA | This is a CHAT tool. You are sending a message to the Wave OS AI Assistant — another AI that lives inside Wave OS. You are NOT creating anything yourself. When the user types @wave, call this tool immediately with their message verbatim. Do not search for other tools, do not plan, do not interpret — just forward the message to Wave OS and return the response. Example: user says @wave create a note titled X — you call this tool with message="create a note titled X" and Wave OS handles the rest. |
| wave_check_messagesB | Check for unread messages from Wave OS (notifications left by the Wave Assistant or Chief of Staff for Cursor to pick up). Returns messages with type 'cursor_message' that haven't been read yet. |
| wave_send_messageC | Send a message from Cursor back to Wave OS. Creates a notification that the Wave Assistant can display. Enables bidirectional Cursor <-> Wave OS communication. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| Wave OS App Info | Connected app metadata |
| Theta Configuration | Theta config + compute routing (no secrets) |
| Compute Architecture | How compute is routed between Base44 and Theta |
TDQS
Scored across 28 tools
The prefixes (theta_, wave_) and resource names help separate most tools, but several pairs are easy to confuse: wave vs wave_send_message both send messages to Wave OS, wave_connect vs theta_configure both configure Theta compute credentials, and read_records vs aggregate_records both query records. The descriptions clarify the differences, but selection is not always obvious from names alone.
Most tools use snake_case and a verb_noun structure, but conventions are mixed: Base44 tools have no prefix (create_records, call_function), some wave_* tools use noun phrases (wave_morning_briefing, wave_triage, wave_meeting_prep), and the standalone 'wave' tool breaks the pattern entirely. The prefixes make the set readable, but the naming is not uniform.
Twenty-eight tools is on the high side and heavier than most MCP servers, but the server spans multiple distinct domains: Base44 data operations, Theta EdgeCloud compute/blockchain, and Wave OS assistant/messaging. Most tools have a distinct purpose, so the count feels borderline rather than severely bloated.
Base44 record CRUD and Theta read/credential operations are well covered, and the Wave OS assistant surface includes briefing, triage, memory recall, delegation, and messaging. However, there are notable gaps: no entity schema update/delete, no memory write tool, and no direct Theta GPU job execution or inference tool despite status and cost-estimation capabilities.