Searches your document collections and returns ranked passages with sources, uploads new files, retires stale ones, and syncs local connectors.
16actions available
Every action runs live through MCP. Nothing to build, nothing to maintain.
Lyro performs semantic search within Needle collections to find the exact document or chunk a customer needs. Results rank by relevance with source references so the answer is both fast and traceable.
Lyro uploads new files to a collection so the team can expand the knowledge base mid-conversation. Future customer queries can now reference the newly added materials.
Lyro retrieves statistics for a collection to show the team how many documents and queries it contains. The team optimizes the collection's scope based on usage patterns.
Each card is one request a support team gets, and the Needle actions Lyro runs to close it.
Lyro runs a natural language query against a named collection, gets ranked results with source references back, and fetches a signed download URL when the full document is needed.
Lyro creates the collection if it does not exist yet, requests signed upload URLs for the content types involved, and adds the files to the collection by URL.
Lyro pages through the files in a collection, deletes the ones superseded by a newer version, and reads back the document count and index size to confirm what is left.
Lyro sets up a connector that watches folders on a device, adds external files to it by URL, and reads back which connectors are configured and what each one is monitoring.
How it works
Connect once, then just ask. There is no workflow builder to learn and nothing to maintain — Lyro reads the Needle actions it has and picks the ones a request needs.
Authorize the Needle account your team already uses — one consent screen, no API keys, no mapping tables. Lyro can only do what you granted that account, and you can disconnect it at any time.
Describe the job the way you would hand it to a teammate. Lyro maps it to the Needle actions that close it and chains as many as the request needs.
The agent runs the actions inside the conversation the customer is already in, so nobody copies data between tabs and your team can take over at any point.
Get started freeNeedle · Lyro
What can you do in Needle without me setting anything up?
These, right away:
Great, let's start with the first one.
FAQ
Setup, permissions, and the limits of what Lyro can do inside Needle.
Yes. Search Collection performs semantic search inside a named collection and returns ranked results with source references attached, so Lyro can point at the document a statement came from rather than asserting it. That is what makes the answer checkable by whoever asked.
All 16 actions your agent can call on Needle, straight from the live MCP connection.
Add files to a collection by providing file URLs.
Add files to a local connector by providing file metadata.
Create a new collection.
Create a local connector that monitors specified folders on a device.
Delete files from a specific collection by providing file IDs.
Delete files from a local connector by filename or file IDs.
Retrieve details for a specific collection by its ID.
Retrieve statistics for a specific collection by its ID.
Get a short-lived signed private download URL for a Needle file.
Get signed URLs for uploading local files to Needle.
Retrieve details of a local connector by its ID.
List all files within a specific collection by its ID.
List collections.
List connectors.
List local connectors.
Perform semantic search within a specific Needle collection and return ranked results with source references.
Same connection, same setup. Pick the next one your team already uses.
Inventories models, clusters, and anomaly detectors, reads their evaluations and predictions, and sets up projects and external data connectors.
Sends prompts to deepseek-chat or deepseek-reasoner with tool calling and thinking mode, and checks balance and model availability first.
Creates a project, reports on its experiments and session history, and deletes the project when the work behind it is finished.
Route a request to a self-hosted model, reuse OpenAI-format endpoints against it, and report which models the host currently has installed.
Creates image, video, LiDAR, and text annotation tasks, batches and finalizes them, reports completion counts, and reassigns Studio projects.
Scores a submitted passage for AI authorship, checks it for plagiarism against online sources, and measures how closely two texts match.

Authorize the account and your agent has all 16 actions from the first conversation.
