RunningHub MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| RUNNINGHUB_DB_PATH | No | Explicit path to the SQLite database. | <data dir>/runninghub.sqlite |
| RUNNINGHUB_DATA_DIR | No | Directory for the local SQLite database and working state. | ~/.runninghub |
| RUNNINGHUB_LIVE_CASES | No | Live harness profile; full is the default and is required for explicit paid live checks. | full |
| RUNNINGHUB_CATALOG_DIR | No | Directory containing the fixed catalog files included in the package. | <cwd>/data/upstream |
| RUNNINGHUB_FFMPEG_PATH | No | FFmpeg executable used for local image previews and video posters. | ffmpeg |
| RUNNINGHUB_PROJECT_ROOT | No | Reserved setting for the projects root. | |
| RUNNINGHUB_WORKFLOW_API_KEY | Yes | Enables the official RunningHub Workflow API adapter. Required for cloud generation. Keep this key secret. |
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 | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
| completions | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| rh_search_modelsC | Search the pinned public RunningHub model catalog. This is discovery only, not account availability. |
| rh_get_model_schemaB | Return a public schema for one catalog endpoint; it does not verify cloud access. |
| rh_estimate_priceC | Estimate a public catalog price. Verify official RunningHub pricing before showing a final cost. |
| rh_validate_payloadB | Validate a payload against a public model schema; this does not execute it. |
| rh_get_integration_guideB | Return public-safe integration guidance for the optional standard model API family. |
| rh_build_example_payloadC | Build an example payload from a public model schema; it does not submit a task. |
| rh_create_workflowB | Create an immutable local workflow revision from an API graph or an empty draft. No cloud task is submitted. |
| rh_get_workflowC | Read a local immutable workflow revision and its validation report. |
| rh_edit_workflowB | Apply an atomic typed graph edit batch using compare-and-swap; failed batches create no revision. |
| rh_validate_workflowB | Return structural, runnable, and backend-compatibility states for a local workflow revision. |
| rh_export_workflowB | Export a local workflow revision as API-format JSON without secrets or cloud submission. |
| rh_projectC | Register, inspect, list, or index a project without changing files outside its explicit roots. |
| rh_assetC | Inspect or register project-owned assets, prepare a tagged provider reference, or explicitly upload one through the configured backend. |
| rh_upload_loraC | Upload one registered project LoRA through the dedicated RHLoraLoader flow; signed upload URLs are never returned or stored as graph references. |
| rh_sceneC | Read, resolve, or upsert normalized scene context and its required asset roles. |
| rh_work_itemB | Create/read/close a requested work item; creation does not create a job or submit a paid task. |
| rh_search_workflowsB | Search the registered workflow library using hard scene requirements; constraints are never weakened to force a result. |
| rh_import_workflowC | Discover API-format workflow JSON files in a registered project or import one explicitly; no provider task is submitted. |
| rh_prepare_generationB | Create an immutable local execution plan from one work item and one workflow revision; no provider task is submitted. |
| rh_run_workflowB | Submit one prepared execution plan through its configured durable backend; repeated request IDs never create a second local submit. |
| rh_jobC | Inspect or advance one durable job without hiding provider errors or retrying an uncertain submit. |
| rh_get_resultsB | Download confirmed provider outputs as validated local original files, render supported images inline in the current MCP chat, and create local preview/poster derivatives without submitting another task. |
| rh_review_resultC | Record one idempotent user review for a saved result; an explicit approved continuation may run one already prepared next plan. |
| rh_get_capabilitiesC | Report local server, catalog provenance, and backend capability evidence without exposing secrets. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 24 tools
Most tools target clearly distinct resources or lifecycle stages, and descriptions explicitly separate local from cloud and discovery from execution. Some potential confusion remains among noun-only multiplexed tools (rh_job, rh_work_item, rh_prepare_generation) and between validate_payload and validate_workflow, but boundaries are generally clear.
All tools use a consistent rh_ snake_case prefix, but the naming pattern is not uniform: many are verb_noun, while project, asset, scene, work_item, and job are noun-only tools that bundle multiple operations. The names are readable but do not follow one predictable convention.
24 tools is heavy for a single MCP server and sits at the upper edge of the 16-25 range before becoming excessive. The domain is broad enough that each tool has some rationale, but the set is large enough to burden selection and increase cognitive load.
The surface covers discovery, workflow creation/read/edit/validate/export/import/search, project and asset handling, scene and work-item context, generation preparation/run/job/results/review, and capability reporting. Some delete or list operations are absent, such as deleting projects/assets or listing jobs, but core generation lifecycle coverage is strong.