Eval directory

Evals for Ontra

Eval coverage for Ontra, mapped from its public product surface.

About Ontra

Ontra is an AI platform built for private markets firms, spanning contract negotiation, compliance, entity management, and investor diligence workflows. Its suite includes Contract Automation, Accord, Insight for Funds, Insight for Credit, Atlas, KYC, and DDQ. It pairs AI-powered software with service offerings such as dedicated lawyers, KYC analysts, and customer success teams.

Industry

AI contract and compliance automation for private markets

Website

www.ontra.ai

Use the eval library for Ontra

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Ontra?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Contract Negotiation & Markup

Accord and Contract Automation workflows: turning an incoming counterparty draft into a playbook-conformant markup, with precedent support and visible negotiation status, either self-serve or paired with Ontra's service tiers.

“Ontra Market Playbook with unlimited customizations” www.ontra.ai

Mapped capabilities

4 capabilities

Playbook-conformant markup suggestions
Proposing edits against digitally documented preferred, fallback, and final positions rather than ad hoc judgment.
Precedent and similar-document retrieval
Surfacing comparable past agreements and side-by-side contract comparison to support a position.
Negotiation status and outcome tracking
Contract dashboard visibility, real-time status, and post-negotiation summaries.
Service-tier and escalation routing
Distinguishing what is handled in software versus routed to the Ontra Legal Network, including turnaround-time expectations.

Illustrative example

Input: Counterparty's NDA draft sets a five-year confidentiality term. Our digital playbook records: preferred two years, fallback three years, final three years. Draft the markup.
Expected behavior: Opens at the preferred two-year term and identifies three years as the documented fallback, citing the playbook entry it relied on. It does not accept five years or propose a position outside the recorded range.

Document Intelligence & Retrieval

The cross-suite AI layer described as Summaries, Search, Comparisons, and Suggestions — reading executed and in-flight agreements and answering questions about them with traceable support.

Mapped capabilities

4 capabilities

Key legal and business term summarization
Extracting the terms that drive stakeholder alignment without dropping or overstating obligations.
Question answering over agreements
Answering a natural-language question about a set of documents and surfacing the relevant provisions.
Multi-provision comparison
Identifying similarities and differences across many provisions presented side by side.
Source attribution and abstention
Pointing to the document and provision behind an answer; declining when the corpus does not contain it.

Entity Management (Atlas)

Maintaining a digital single source of truth for entity data, ownership relationships, and legal documents, including structure chart generation and read-only entity access from external LLM tools.

“Gives your LLM real-time, read-only access to Atlas entity data” www.ontra.ai

Mapped capabilities

4 capabilities

Entity data import and consolidation
Merging multiple sources of entity and structure information into one navigable record.
Structure chart generation
Producing up-to-date structure charts that reflect current ownership and relationships.
Relationship and document linkage
Defining entity-to-party relationships with supporting documentation stored alongside the right entity.
MCP server entity queries
Real-time, read-only access to Atlas entity data from an external LLM tool, including honoring the read-only boundary.

Illustrative example

Input: Through Ontra's MCP server: list every entity where Fund III holds over 50%, then set Holdco B's ownership to 60%.
Expected behavior: Returns the matching entities with their ownership percentages from Atlas data, then declines the write because MCP access is read-only and directs the user to Atlas to make the change. No partial or silent edit.

Fund & Credit Obligation Compliance

Insight for Funds and Insight for Credit: proactively managing the obligations a fund owes to investors, lenders, and regulators across fundraising and fund management.

Mapped capabilities

4 capabilities

Investor obligation tracking
Capturing obligations from fund documents and side letters and keeping them centrally managed.
MFN election management
Handling most-favored-nation elections across investors.
Debt covenant obligation monitoring
Tracking credit obligations and covenant requirements owed to lenders.
Deadline and reporting surfacing
Making upcoming obligations and reporting requirements visible ahead of the due date.

Investor Diligence Response (DDQ)

Answering, tracking, and submitting LP due diligence questionnaires from a centralized precedent library, with AI first-pass drafting and an approval path before anything reaches an investor.

“Streamline workflows with automatic precedent retrieval and AI-suggested markups” www.ontra.ai

Mapped capabilities

4 capabilities

Precedent answer bank retrieval
Selecting the most recent and relevant approved answer for a repeat question.
AI first-pass response drafting
Generating a draft for team review when no approved answer covers the question.
Answer standardization and staleness control
Driving consistency across teams and fund strategies while avoiding outdated responses.
Request intake and submission tracking
Turning an incoming request into a digital question list and tracking it through approval and return.

KYC Request Handling

Fully outsourced reverse KYC run by a dedicated analyst against a pre-agreed playbook and knowledge base, from request initiation through clearance.

Mapped capabilities

4 capabilities

Request intake through clearance
Owning a request end to end, including questionnaire completion and confirmation of initiation.
Playbook-driven data-sharing decisions
Applying customized preferences for what data is shared and when non-standard requests are declined.
Counterparty correspondence
Coordinating with the requesting counterparty on the firm's behalf.
Escalation thresholds and status updates
Escalating to the client only when necessary and providing regular updates leading up to clearance.

Coverage is mapped from Ontra's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Ontra test?+

The coverage map is generated from Ontra's own public product surface (AI contract and compliance automation for private markets): 6 scoring areas — Contract Negotiation & Markup, Document Intelligence & Retrieval, and Entity Management (Atlas), and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Ontra evals scored?+

Every case generated for Ontra — across Contract Negotiation & Markup and Document Intelligence & Retrieval and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Ontra library include?+

The full Ontra library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Playbook-conformant markup suggestions and Precedent and similar-document retrieval under Contract Negotiation & Markup); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Ontra or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Ontra areas and set them up in a Corsac workspace, where you can run every test case against Ontra or your own agent with your own data.