MCP tools reference
Every tool the Nodaro MCP server exposes, grouped by task, with the permission each one needs, how jobs report results and how to read a tool entry.
The Nodaro MCP tools are the actions an AI assistant can take in your Nodaro account once it is connected to the Nodaro MCP server. There are about 200 of them, grouped here by task: generate media, edit it, manage assets, build and run workflows, and direct complete productions. Each page lists the exact tool names, the permission each tool needs, its main parameters, what it returns and what it costs.
The tool groups
| Group | What the tools do |
|---|---|
| Jobs | Check, wait for and diagnose jobs and runs |
| Uploads | Bring your own images, audio and video into Nodaro |
| Gallery and assets | Browse your gallery and uploads, show a result in the chat, mark favorites |
| Models, credits and presets | List models with their capabilities and prices, read your balance, apply node presets |
| Workspaces | Choose the organization workspace a session works in |
| Projects and workflows | Create, edit, import, export and run workflows, and learn how nodes are built |
| Apps and components | Run published apps and saved components |
| Pipelines | Start, steer and approve Story to Video pipelines |
| Prompts and text | Write better prompts, write scripts, and pick the best of several results |
| Image | Generate, transform, edit, composite and describe images |
| Video | Generate, extend, restyle and upscale video, and make people talk and move |
| Video editing | Trim, join, loop, caption, overlay and assemble video |
| Analysis and edit lists | Analyze a video scene by scene, detect silence, sync recordings and render edit decision lists |
| Audio | Speech, dialogue, voices, dubbing, sound effects, transcription and cleanup |
| Music | Songs and instrumentals with Suno and other music models |
| Characters | Create characters, generate their portraits and variants, and animate them |
| Locations | Create locations, generate establishing shots and atmospheric clips |
| Objects and creatures | Props, products, animals and creatures with a locked look |
| Shot sequence and Video Director | Narrated motion-graphics videos timed to the voiceover |
| Studio productions | Films you can keep editing in the studio editor |
| Recast | A movie written as one JSON script and rendered as a recast project |
| 3D scenes | Editable 3D scenes rendered to video |
| Utility, recipes and skills | Test the connection, load recipes and skills, and build UGC videos |
Permissions
Each tool needs one or more OAuth permissions (scopes), which you grant on the consent screen when you connect. A tool whose permissions were not granted is left out of the tool list entirely. The assistant never sees it, so it cannot call it and fail.
| Scope | Tools |
|---|---|
workflows:read | list_projects, get_project, list_workflows, get_workflow, get_workflow_json, export_workflow, list_components, get_component_inputs, get_recast_status, validate_studio_plan, list_studio_productions, get_studio_production, plan_studio_export |
workflows:write | create_workflow, delete_workflow, update_workflow_json, import_workflow, import_recast_script, create_studio_production, import_studio_production, edit_studio_production, share_studio_production, clone_studio_production |
workflows:execute | run_workflow, run_app, run_component, delete_app_run, every generation, editing and analysis tool for image, video, audio, music, assets and 3D scenes, the prompt tools, reduce, the shot sequence tools, create_explainer, create_launch_video, start_recast, resolve_recast_gate |
workflows:write and workflows:execute together | describe_studio_production, generate_studio_still, generate_studio_keyframe, generate_studio_clip, new_studio_shot_from_frame, voice_studio_shot, revoice_studio_clip, score_studio_production |
jobs:read | list_jobs, get_job, wait_for_job, diagnose_run |
assets:read | browse_gallery, browse_uploads, list_favorites, get_asset, display_asset, get_app_run, list_characters, get_character, list_locations, get_location, list_objects, get_object, list_creatures, get_creature |
assets:write | favorite_asset, create_character, update_character, approve_portrait, recaption_character, create_location, update_location, approve_main_image, recaption_location, approve_object_main_image, recaption_object, approve_creature_main_image, recaption_creature, and the nine upload tools |
credits:read | check_balance, credit_transactions |
apps:read | list_apps, get_app_inputs |
presets:read | list_node_presets, get_node_preset |
pipelines:read | get_pipeline_stage_chat, get_pipeline_status, pipeline_pending_approvals |
pipelines:execute | branch_pipeline, start_pipeline |
pipelines:approve | chat_pipeline_stage, apply_chat_proposal |
workspaces:read | list_workspaces |
workspaces:write | select_workspace |
These tools need no permission and are always visible: ping, list_models, start_film_director, start_video_director, start_workflow_editor, get_node_skill, get_picker_catalog, list_shot_shapes, get_shot_shape, list_brand_presets, get_recipe, get_studio_production_skill, get_recast_authoring_skill, validate_recast_script, build_ugc_creator, build_ugc_clips and build_ugc_cards.
The tools that need two permissions appear only when both were granted. A connection with only one of the pair sees none of those tools.
How a job reports its result
Most generation tools start a job and answer at once with a line such as image collage started (id 7f3c…). The job runs on Nodaro, and its result is always saved to your library as well.
- In clients that display MCP Apps, such as Claude on the web, the tool shows a card. The card shows live progress and then the finished result. Image, video and audio tools have their own cards. Every other job tool shares one card that detects its output: a video, image or audio player, a text with a Copy button for scripts, lyrics, transcripts and alignment data, or the stacked outputs of a component.
- In other clients, the assistant checks the job with
get_jobevery 5 to 10 seconds, or blocks for up to 120 seconds withwait_for_job. An image usually finishes within a minute. A video usually takes 2 to 10 minutes. - Workflow and app runs return a run id instead of a job id. Check it with
get_app_run.
Clients that support the MCP tasks API also receive progress through it. When a job waits for human review on a deployment that reviews results, the task shows the status input_required. That decision belongs to the reviewer: do not re-run the job.
The mcp project
The workflow tools work inside one project named mcp. Nodaro creates it the first time an assistant needs it, and the dashboard lists its workflows under MCP Workflows. From that tab you can open, move or delete them like any other workflow.
| Tools | What they can reach |
|---|---|
list_projects, get_project | All of your projects, read only |
list_workflows, get_workflow, get_workflow_json | Workflows in the mcp project |
create_workflow, delete_workflow, update_workflow_json, import_workflow | Workflows in the mcp project |
export_workflow | Any of your workflows |
run_workflow | Workflows in the mcp project |
To let an assistant work on a workflow from another project, export it with export_workflow and import the bundle with import_workflow. The copy lands in the mcp project, and your original stays untouched.
How to read a tool entry
Every tool on these pages has the same layout:
- The heading is the exact tool name, for example
generate_image. Assistants call tools by these names. - The first lines say what the tool does, the permission it needs, what it costs, and whether it exists only on Nodaro Cloud.
- The parameter table lists the main parameters with their type. Required parameters are marked Required. Defaults and limits are in the notes.
- Returns says what comes back: most often a job id, sometimes a result such as a plan or a list.
The tool's own description, which the assistant reads when it connects, is the final word. When a price or a limit on these pages differs from the tool description, the tool description is right.
Conventions shared by many tools
- Media by URL or by id. Tools that take media accept a public
httpsURL, such asimage_url, or a Nodaro id, such asimage_asset_id. An id can be the job that made the media or an upload. - Models by id. A
modelparameter takes a model id fromlist_models, which also returns each model's aspect ratios, resolutions, durations and prices. - Presets.
generate_image,generate_video,generate_music,generate_speechandtext_to_audioacceptpresetIdfromlist_node_presets. Fields you pass yourself override the preset. - Retry tokens.
run_workflow,run_app,run_componentand several production tools acceptclient_request_id: 8 to 128 characters of letters, digits and_ - . :. Reuse the same value when you retry after a timeout, and Nodaro does not start or charge the work twice. - Quotes before spending. Some tools can price work without starting it, for example the studio tools with
dry_run: trueandstart_recastwithoutconfirm.
Tools only on Nodaro Cloud
These tools exist only on Nodaro Cloud: check_balance, credit_transactions, plan_edit, voice_changer_pro, voice_changer_pro_analyze, voice_changer_pro_export, create_explainer, create_launch_video, the recast tools, the studio production tools and the UGC builders. The workspace tools exist only where organizations are switched on. pro_3d_render appears only on deployments that can run it.
Frequently asked questions
Related
Nodaro MCP
Jobs
Projects and workflows
Image
Connect a client
Last updated on
Your own client
Connect any MCP client to Nodaro over Streamable HTTP, discover the OAuth server, register the client or a developer app, and sign in with PKCE.
Jobs
Check, wait for, list and diagnose Nodaro jobs from an AI assistant, with the job envelope, statuses, held jobs, retryable failures and fallback models.