--- title: "Coveo Integration" description: "Put Coveo enterprise search behind your AI agents. Agents run relevance-ranked queries and query suggestions, manage content sources, and push, update, or delete documents in Push sources for indexing." url: https://flowrunner.ai/integrations/coveo date_modified: 2026-08-01T02:40:32-07:00 --- # Coveo [Vector Stores & AI Infra](https://flowrunner.ai/integrations/category/vector-stores-ai-infra) Put Coveo enterprise search behind your AI agents. Agents run relevance-ranked queries and query suggestions, manage content sources, and push, update, or delete documents in Push sources for indexing. [Verified](https://flowrunner.ai/integrations/verified "What does verified mean?") · 9 actions · API key · available [Coveo website](https://coveo.com/) · [Platform Documentation](https://docs.coveo.com/) · Capability data verified 2026-07-27 1. The quarterly content refresh kicks off with the list of new, revised, and retired articles 2. Set Source Status marks the Push source for an incremental refresh so progress reports correctly 3. Batch Add or Update Documents indexes the new and revised articles with their fields 4. Search runs the team's benchmark queries to confirm the fresh content is retrievable and ranking 5. The knowledge manager approves the removal list before anything leaves the index 6. Batch Delete Documents removes the retired items, and Set Source Status returns to Idle 7. The refresh report posts with counts indexed, removed, and benchmark results ## What This Integration Enables Most AI retrieval pipelines quietly rebuild what the search team already built: ranking, filtering, permissions, synonyms. Coveo is the case for not doing that. The Search action runs agent queries through the organization's tuned query pipelines, respecting the permissions of the key that asks, so an agent's retrieval step inherits years of relevance work instead of starting from cosine similarity. And the Push source actions close the loop from the other side: agents keep the index current as content changes, so what search returns is what is actually true. - Power agent retrieval with relevance-ranked Search across enterprise knowledge, permissions included - Narrow queries with advanced and constant query expressions, facets, and paging - Serve typeahead with machine-learning query suggestions from trained pipelines - Push, update, and delete documents in Push sources, one at a time or in batches - Wrap bulk runs with Set Source Status so indexing progress reports accurately - Inspect sources and their document counts before writing anything ### Without FlowRunner **Agents retrieve from whatever is lying around**: The AI answer pipeline searches a stale export because the real index has no API in the loop **Indexing is an IT ticket**: New content becomes findable whenever the next crawl happens, which nobody can predict **Search quality degrades silently**: Retired products and old policies keep ranking because nothing ever removes them ### With FlowRunner **Retrieval runs through governed relevance**: Agents query the same pipelines, ranking rules, and permissions the search team already tuned **Content is searchable the hour it ships**: Publish events push documents into the index directly, wrapped in proper source status **The index has a maintenance loop**: Refresh cycles add, verify, and remove on schedule, with a person approving what disappears ## Use Case Scenarios ### Published means findable, same hour When an article ships in [Contentful](https://flowrunner.ai/integrations/contentful), the flow resolves the Push source with List Sources and calls Add or Update Document with the article body and its metadata fields, keyed on the article's URI. Once Coveo processes the push, the piece is searchable in every experience the index serves. A follow-up Search for the article's target phrase confirms retrievability, and a miss files a task instead of leaving the gap for a customer to find. ### The agent that answers from tuned search, not a side index A question arrives in [Slack](https://flowrunner.ai/integrations/slack). The flow runs Search against the support hub's query pipeline, pulling the top results with their excerpts, and hands them to the answering model as grounding. The results respect the asking key's permissions, so the agent cannot leak content its audience should not see, and the ranking reflects the relevance rules the search team maintains. The drafted answer posts with its sources; a summary page lands in [Notion](https://flowrunner.ai/integrations/notion) for recurring questions. ### The bulk load that reports like a real crawler A migration moves a policy library from file storage into search. The flow pulls documents from [Google Drive](https://flowrunner.ai/integrations/google-drive-api), calls Set Source Status with Rebuild, pushes the collection through Batch Add or Update Documents with a valid URI per document, and sets the source back to Idle when done. Get Source confirms the document count matches expectations, and the discrepancy report goes to the team before anyone declares the migration finished. ## Human-in-Loop Highlight Batch Delete Documents has a cascade option: give it a URI and it can remove every child item whose Document ID begins with that prefix. That is the right tool for retiring a product's entire documentation tree, and it is also one wrong prefix away from silently removing a whole section of the knowledge base from search for every employee and customer at once. Nothing errors when search coverage disappears; people just stop finding things and assume they never existed. So FlowRunner refresh flows never let a computed deletion list execute itself. The agent assembles the candidates, expands what each cascade prefix will actually match, and posts the full manifest: these 87 documents, these three parent URIs, this many children under each. The knowledge manager reads the manifest, catches the prefix that would have taken the active integrations guide with it, and approves the rest. Deletion runs at batch speed only after a person has seen exactly what will vanish from search. Agent processes routinely Detects exception requiring judgment Clear match Continues automatically Ambiguous Routes to human via preferred channel Human decides Agent resumes with decision ## Agent Capabilities 9 actions ### Search 2 - **Search** Runs a query against a Coveo index through the Search API v2, returning ranked results, total count, facets, and query metadata. Supports keyword expressions, advanced and constant query filters, paging, and targeting a specific search hub or pipeline so configured relevance rules apply. Results respect the permissions of the API key used. - **Get Query Suggestions** Returns query completions for a partial input, powered by Coveo Machine Learning when a Query Suggestions model is configured on the target pipeline. The building block for typeahead experiences, scoped by search hub and pipeline. ### Sources 2 - **List Sources** Lists the content sources configured in the organization, including the Push sources agents index into, with each source's ID, name, type, and status. Requires source management privileges. - **Get Source** Retrieves a single source's full configuration and current status, including document count and mappings. The verification step after bulk loads. ### Document Management 5 - **Add or Update Document** Pushes a single item into a Push source, creating or replacing it by its Document ID, which must be a valid unique URI. Carries indexable text plus additional index fields. The publish-to-searchable step for event-driven indexing. - **Delete Document** Deletes a single item from a Push source by Document ID, optionally cascade-deleting child items under the URI. Removal takes effect once Coveo processes the operation. - **Batch Add or Update Documents** Pushes multiple items into a Push source in one operation, each with a unique URI, title, and text body plus optional fields. The bulk path for migrations and scheduled refreshes. - **Batch Delete Documents** Deletes multiple items in one operation by Document ID, optionally cascading to children. The high-leverage removal this page's human gate exists for. - **Set Source Status** Updates a Push source's operational status around bulk runs: Rebuild before a full re-push, Refresh or Incremental before an update pass, Idle when the run completes. The discipline that keeps indexing progress honest. ## Frequently Asked Questions ### What can FlowRunner do with Coveo? FlowRunner agents can run Search, Get Query Suggestions, and List Sources in Coveo, plus 6 more actions. ### Does connecting Coveo to FlowRunner require OAuth? No. Coveo connects to FlowRunner with an API key, no OAuth flow required. ### Can Coveo trigger a FlowRunner workflow automatically? Coveo doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger. **Work at Coveo?** This integration exposes Coveo to AI agents on every FlowRunner plan, including through MCP, at no cost to you. [See what FlowRunner offers integration partners](https://flowrunner.ai/integrations/partners), including how to keep this page current. --- Markdown version of https://flowrunner.ai/integrations/coveo. Site index: https://flowrunner.ai/llms.txt