Snack Prompt
AIStore and share prompts, snippets, and documents through the Snack Prompt Integration API. Agents fetch the approved prompt version rather than one copied into a flow by hand.
What This Integration Enables
Prompts are configuration that most teams still treat as text. They get pasted into a flow, tuned by whoever was closest to the problem that week, and copied into the next flow that needs something similar. Six months later nobody can say which version is live, why it was changed, or who agreed to it. Snack Prompt puts that text somewhere it can be found and versioned. Elementals are prompts, snippets and documents, their templates carry variable tags written with a hash prefix such as #topic or #style, and Prepare Template resolves those tags with your values at runtime and returns finished prompt text. Get Template Tags reports exactly which variables a template expects, so a flow can build its inputs from the template rather than hard-coding them and hoping they still match.
The consequence is what makes this a governance connector rather than a storage one. If the wording lives in Snack Prompt and the flow resolves it at runtime, then the prompt in production is whatever was last published, with no deploy in between. That is a real benefit and a real obligation, which is the same balance FlowRunner's Code of Conduct draws around agent behaviour generally. The publication step is where human-in-the-loop belongs on this connector, because it is the only point at which a person can still choose.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Resolving the approved prompt at runtime
A classification flow no longer carries its instruction text. At the point where it used to hold a paste, it calls Get Template Tags to see the variables the stored template expects, fills them from the record in hand, and calls Prepare Template to get the finished prompt back. The resolved text goes to the model. When the wording needs to change, it changes in one place, and every flow that resolves it gets the new version without anyone editing a flow.
Compiling a curated list into a knowledge pack
A team has assembled a list of the prompts and reference documents that define how they write. Rather than exporting them one by one, the agent calls Compile List Knowledge Base, which turns the whole list into a single export as plain text, Markdown or structured JSON. That pack loads straight into a Pinecone index, so a retrieval step answers from the team's own approved material rather than from whatever the model absorbed during training.
Pulling a themed working set into a campaign flow
Ahead of a launch, the marketing team tags every prompt that belongs to the campaign. When the flow runs, the agent calls List Tags to resolve the tag name to an ID, then Get Tag Items to pull every prompt carrying that tag in one step. The campaign flow works from a set the team curated, and adding a prompt to the campaign is a tagging action in Snack Prompt rather than a change request against the workflow.
Human-in-Loop Highlight
Update Elemental is the operation to gate, because it is a deploy that does not look like one. The instant a revision is published, every flow calling Prepare Template resolves the new wording on its next run. There is no staging environment between the edit and production, and there is no rollback beyond publishing the old text again from memory. Two details sharpen the risk. The API requires title, type, template and category on every update, so anything the caller does not resend is overwritten rather than left alone. And visibility on an elemental can be Public, meaning visible to everyone on the platform, which cannot be taken back once somebody has read it. So the agent prepares the evidence and stops: "Proposed revision to the invoice classification prompt. Current and proposed wording attached, both resolved with the same three sample invoices, outputs paired. The proposal drops the #currency tag, which two live flows still pass in. Visibility stays Unlisted. Publish, or send back?" The owner approves, and the record of who approved which wording exists because a person was actually in the path. The same gate covers Delete Elemental and Delete Multiple Elementals, which are permanent and quietly break every list that referenced them.
Agent Capabilities
30 actionsElementals
10- Search Elementals Searches the whole platform for elementals, which are prompts, snippets and documents, matching a query across title, description and template content. Results are paginated and orderable. Used to discover community prompts before adopting one into a flow.
- List My Elementals Retrieves the authenticated user's own elementals of a given kind, with pagination. Used to enumerate a private library, as opposed to searching the whole platform.
- Get My Elementals Summary Returns a compact key and value summary of the user's elementals of a given kind without the full record for each. Used to build a lightweight picker or to check the size of a library before paging through it.
- Get Elemental Retrieves a single elemental by ID, returning its template, description, type and metadata, rendered as plain text, Markdown or structured JSON and narrowable to chosen fields. Used to read the current wording before proposing a change.
- Create Elemental Creates a prompt, snippet or document in your library. Titles run 10 to 100 characters, a category is always required, and prompts additionally require topics. Prompt templates may embed variable tags with a hash prefix that Prepare Template later fills in.
- Update Elemental Updates an existing elemental you own. The API requires title, type, template and category on every update, so current values must be supplied for anything not being changed. Used to publish an approved revision from a flow, after review.
- Delete Elemental Permanently deletes an elemental you own. This cannot be undone, and any list that referenced it loses the entry.
- Delete Multiple Elementals Permanently deletes several elementals you own in one request. Deletion cannot be undone, and the call fails for any ID you do not have edit rights over.
- Favorite Elemental Adds an elemental to the user's favorites, which is how a community prompt gets bookmarked into a personal working set.
- Unfavorite Elemental Removes an elemental from the user's favorites. The elemental itself is untouched and only the bookmark is dropped.
Templates
2- Get Template Tags Lists every variable tag embedded in an elemental's template, such as #name, #topic or #style. Used before Prepare Template so a flow builds its inputs from the template's own declaration rather than a hard-coded guess.
- Prepare Template Renders an elemental's template with your values substituted for its variable tags and returns finished prompt text ready to send to a model. This is the operation that makes Snack Prompt a prompt store rather than a document library.
Knowledge Base
2- Compile Elemental Knowledge Base Compiles a single elemental into an export as plain text, Markdown or structured JSON, optionally narrowed to chosen fields. Used to feed one prompt or document into a retrieval index without reconstructing its formatting.
- Compile List Knowledge Base Compiles every elemental in a list into a single export as plain text, Markdown or structured JSON. Used to turn a curated list into a self-contained knowledge pack in one step.
Lists
7- Search Lists Searches the whole platform for lists, which are curated collections of elementals, with paginated results. Used to find a shared collection worth compiling.
- List My Lists Retrieves all lists belonging to the authenticated user. Lists group related prompts, snippets and documents and are the unit that knowledge base compilation operates on.
- Get List Retrieves a single list by ID, optionally expanding every elemental it contains rather than returning identifiers only, with a chosen output format. Used when the whole collection is small enough to pull at once.
- Get List Items Retrieves the elementals inside a list page by page. Preferred over expanding a large list, since it avoids pulling the entire collection into one response.
- Create List Creates a list for organizing elementals, optionally seeded with an initial set of items. Used to assemble a reusable knowledge pack from prompts that already exist.
- Update List Updates an existing list you own. The title is required on every update, and supplying item IDs sets the list's membership. Used to refresh a curated collection from a flow.
- Remove Elementals From List Removes specific elementals from a list without deleting the elementals themselves. Used to prune a collection while leaving the underlying prompts intact.
Tags
4- Create Tag Creates a tag for categorizing elementals of a given type, with a display colour. Tags are personal labels, separate from the platform's categories and topics.
- List Tags Retrieves every tag available to the user, including system tags as well as ones you created. Used to resolve a tag name to the ID a lookup needs.
- Get Tag Items Retrieves every elemental carrying a given tag. Used to pull a themed working set, such as every prompt belonging to one campaign, into a flow in a single step.
- Delete Tag Deletes a tag you created. The elementals it was applied to are unaffected and only the label is removed.
Reference Data
4- List Elemental Types Retrieves the elemental types the platform supports, currently Prompt, Snippet and Document, with their numeric IDs. Used when a type ID is required rather than a friendly name.
- List Categories Retrieves the categories available for a given elemental type. A category is required when creating or updating an elemental and the valid set differs per type.
- List Topics Retrieves the topics available within a category. Topics are required when creating a prompt, so this resolves the second half of the identifiers creation expects.
- List Visibility Types Retrieves the visibility settings the platform supports, Public meaning visible to everyone and Unlisted meaning accessible only via a direct link, with their numeric IDs. Worth reading deliberately before anything is published.
Account
1- Validate API Token Validates the configured API key and returns the authenticated user's ID, name, username, email and avatar. Used as a connection health check before a longer sequence of calls.
Frequently Asked Questions
What can FlowRunner do with Snack Prompt?
FlowRunner agents can run Search Elementals, List My Elementals, and Get My Elementals Summary in Snack Prompt, plus 27 more actions.
Does connecting Snack Prompt to FlowRunner require OAuth?
No. Snack Prompt connects to FlowRunner with an API key, no OAuth flow required.
Can Snack Prompt trigger a FlowRunner workflow automatically?
Snack Prompt doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Snack Prompt
$100 in credits. No card required. Connect in minutes.