.svg)
Claude Connector for eDiscovery & AI Data Collection
Collect Claude conversations, files, and shared artifacts for eDiscovery, compliance, and investigations, without waiting on IT. The Onna + Claude Connector delivers defensible, structured access to AI-generated content across your organization.
Why Connect Claude To Your eDiscovery Platform
Claude is quickly becoming a core layer of enterprise productivity, powering AI-generated content, decision support, and internal workflows. That introduces a new category of discoverable data: AI interactions.
Collecting Claude data isn’t straightforward:
AI conversations are not stored like traditional files
Content is generated dynamically, prompt to response
Access is governed by Anthropic’s Compliance API
Conversations hold the breadcrumbs to every related file and artifact
The Onna + Claude Connector solves this by using Anthropic’s Compliance API to collect AI conversations in a structured, defensible format. This ensures organizations can preserve, review, and produce AI-generated content alongside the rest of their collaboration data.
Built for AI Data Governance
at Scale
The Onna + Claude Connector is purpose-built for AI data governance at scale. Key capabilities include:
Secure authorized connection via a scoped API key
Organization and custodian discovery and selection
Custodian-based collection of Claude activity
Audit logs for all collection activity
One-time and recurring sync modes
Incremental sync using time-based checkpoints
Resumable workflows for long-running collections
Structured conversation ingestion (prompt and response pairs)
Metadata capture including model version and timestamps
These capabilities let organizations operationalize AI governance workflows across legal, compliance, and security teams.
What Data Can Be Collected from Claude
The connector captures structured AI interaction data from Claude environments via Anthropic’s Compliance API, including:
Conversations
Full Claude conversation threads
Chronological prompt and response sequences
Custodian-linked conversation grouping
Messages (per turn)
User prompts (input to Claude)
AI-generated responses
Parent-child relationships between prompts and responses
Files and artifacts
User-uploaded files and Claude-generated files, collected as child resources of the conversation
Shared artifacts generated in Claude
Participants
Custodian (organization user)
Claude system (synthetic AI participant)
Content Structure
Conversation threads as standalone reviewable units
Clearly distinguishable human vs. AI interactions
This structured model ensures investigators can fully reconstruct AI-assisted decision-making workflows.
Claude Metadata Collected
In addition to standard metadata, the connector captures AI-specific attributes:
Conversation and thread IDs
Custodian identity
Organization
Message-level timestamps
Conversation timestamps (created and modified)
Model version used for each response
Thread structure and conversation grouping
This metadata enables advanced filtering, timeline reconstruction, and model-level analysis across AI usage.
Onna + Claude
Connector Requirements
To connect Claude to the Onna platform, the following requirements must be met:
Claude Enterprise with Compliance API access enabled
An administrator who can generate a scoped Compliance API key
Organization admin permissions
Authorized connection configured in Onna
Because Claude data is accessed through Anthropic’s Compliance API, only content available through that API is retrievable for collection.
How Claude Data Collection Works
The connector uses an API-driven workflow powered by Anthropic’s Compliance API.
Add Claude as a data source.
Configure the data source in Onna with a name and sync settings.
Authenticate the connection
Add a scoped Compliance API key; Onna validates it against the organizations endpoint.
Select organizations.
Retrieve and select one or many organizations the key can reach. “Select all” syncs new organizations going forward.
Select custodians and data types.
Choose the custodians in scope and the optional data to collect (user files, Claude-generated files, artifacts). Conversations are always collected.
Initiate collection.
Onna requests each custodian’s conversations and associated files through the Compliance API.
Ingest and normalize data
Onna downloads, parses, and structures conversations into searchable records organized by organization, custodian, and resource.
Search and review
Collected Claude data is immediately searchable within the platform and ready to cull and export.
Claude Data Collection Options
The connector supports flexible sync configurations:
One-Time Sync
Targeted collection for investigations or legal matters.
Auto Sync
Continuously captures new Claude conversations and updates using time-based checkpoints.
Auto Sync + Archive
Maintains a defensible long-term archive of AI interactions.
Common Claude eDiscovery Use Cases
AI Governance & Oversight
Monitor how employees use generative AI across the organization.
Litigation Response
Collect AI-generated content relevant to legal matters and case strategy.
Internal Investigations
Analyze prompts and outputs tied to policy violations or misconduct.
Regulatory Compliance
Ensure AI usage aligns with emerging regulatory requirements.
%201.webp){kind=logo}
%201.webp)
Onna + Claude Connector FAQs
Yes. The connector collects through Anthropic’s Compliance API using a scoped API key.
Conversations, user prompts and Claude responses, user-uploaded and Claude-generated files, shared artifacts, and associated metadata.
The connector supports one-time and recurring sync with time-based checkpoints; it is not real-time streaming.
Yes. Collection is custodian-based and can be scoped by organization, custodian, and date range.
Onna collects whatever the Compliance API serves at collection time. If a chat was deleted by a user or admin but is still within Anthropic’s retention window, Onna captures the available high-level metadata and generates a stubbed record in the custodian’s folder, marked as “deleted in the origin.” The content itself isn’t recoverable, but the flagged placeholder preserves a defensible indication that a conversation existed and was removed, which teams can investigate further. Recurring sync and early preservation remain the best way to capture content before it’s deleted.
Yes. User-uploaded and Claude-generated files are collected as child resources of the conversation, with shared artifacts.
Yes. Collection runs with full audit logs and a complete chain of custody.
Start Collecting Claude Data for eDiscovery
Connect Claude in minutes and begin collecting AI-generated content across your organization for compliance, investigations, and governance.