Metal
Metal, made by New York based Metal Technologies Inc, is an AI context platform for private equity and private credit firms. Its Context Graph pulls a firm's deal rooms, file shares, email and CRM records into one structure, resolves the companies, people, documents, figures and investment decisions in them, and ranks competing figures by source authority and recency while keeping alternatives visible. A figure can point back to its source document and page where the connector supports citation, and permissions follow the source systems.
Teams run Workflows on the graph to screen deals, analyze source material, prepare investment committee work and monitor portfolios. Skill Builder turns a firm's own methods into reusable workflows, and a Meeting Transcriber adds call notes. An open API and a native Model Context Protocol server pass the context to Claude, ChatGPT, Gemini, Copilot or a firm's own agents. Connectors cover DealCloud, SharePoint, Egnyte, Outlook and DocuSign, a PitchBook partnership adds market data, and the platform supports CRM writeback with audit logging for sensitive changes.
Metal runs each firm in its own tenant and says client data is never used for training by Metal or any AI provider. Its public subprocessor list names OpenAI, Anthropic, Reducto and Fireworks as AI providers, with AWS hosting in the US. In its own test of 51 paired questions, Metal scored 95.1 against 86.6 for a leading enterprise file platform and answered in half the time. Clearlake Capital, Berkshire Partners, Valesco Industries and Blue Wolf Capital use it. Taylor Lowe, Sergio Prada and James O'Dwyer founded the company, which is backed by Y Combinator, Swift VC and Base10.
Capability Axes
Capability grades
15 of 15 axes rated · 10 graded A or B
The product is the AI layer: entity resolution, source ranking and continuous ingestion build the Context Graph, and workflows, skills and outside AI tools run on it. Without the models there is no graph to query and no diligence output to sell.
Workflows run screening and diligence tasks on the graph, surface competing values with their sources for an analyst to resolve and can write findings back to the CRM, with sensitive writes audited. Firms build their own skills with Skill Builder. Which actions need approval before they run, across workflows and connectors, is not described.
Figures trace to source documents and pages where the source supports citation, and competing values stay visible, ranked by source priority and recency. Metal's own evaluation on 51 paired questions scored its answers 95.1 against 86.6 for a leading enterprise file platform, with 93% fewer errors per answer than structured workflows. Validation data for the ranking, error rates by source type and a model version history are not published.
Berkshire Partners' chief technology officer Eric Souza says Metal fast tracked the firm's use of AI, and the firm uses it to process expert transcripts and find similar companies. Clearlake Capital's Tony La Rosa and Valesco Industries' Patrick Floeck are also quoted. Metal's own 51 question evaluation reports a 95.1 quality score against 86.6, answers in 3.8 minutes rather than 7.6 and 41% fewer tokens, and no customer case pairs a named firm with a measured result.
Each firm runs in its own tenant, and Metal states that client data is never used for training, not by Metal and not by any AI provider. Customer data is deleted when a client leaves, and a full export is delivered within 30 days of a request. Retention periods at the named AI providers for model calls are not stated.
Personal data handled through the service falls under Metal's privacy policy, and the trust center adds the full subprocessor list with locations, controls for data retention, classification and deletion on exit, and GDPR compliance. The customer data processing agreement and any GLBA service provider terms are not public.
Metal's trust center at security.metal.ai shows SOC 2 Type II across security, availability, processing integrity, confidentiality and privacy, with a control list covering penetration testing, encryption, access and recovery. The company announced its first SOC 2 audit, by Advantage Partners, in 2024, and ISO 27001 is in progress. The trust center lists the attestation and its controls, but no report opens from it.
Metal sells research and diligence software to investment firms, whose analysts review the sources and outputs before acting. The terms present it as software, and the company's regulatory position, including whether any advisory registration applies, is not stated.
Source ranking, visible conflicts between values, permissions that follow the source systems and audited sensitive writes give a deal team ways to review what the graph produces. Those are product controls rather than an AI governance program, and no fairness evaluation, model governance process or responsible AI policy is described. The bias surface is modest, since the product supports research on companies and deals rather than decisions about individuals.
The graph shows the sources behind each figure, keeps competing values visible and logs sensitive changes written to a connected CRM, which lets a deal team trace and fix a wrong answer. The website terms limit liability for use of the site, and the customer agreement that would allocate responsibility for AI outputs is not public.
Metal's public subprocessor list names every party in the chain: OpenAI, Anthropic, Reducto and Fireworks as AI service providers, Amazon Web Services for cloud hosting, MongoDB Atlas for storage, Vespa.ai for search, Datadog for observability, PostHog for analytics and PropelAuth for authentication, all in the US. Claude, ChatGPT, Gemini and Copilot also appear as tools that can draw on the graph through the MCP server, which is access rather than supply. Which provider serves which workflow, and how model versions change, are not described.
Connectors bring in DealCloud, SharePoint, Egnyte, Outlook and DocuSign records, and the PitchBook partnership adds market and deal data to the workspace. The open API exposes companies, people and deals, the MCP server connects the graph to outside AI tools with per user OAuth or service keys, and findings can be written back to the CRM.
Each firm runs in its own logically isolated tenant, encrypted in transit and at rest, and the security page says data is processed and stored within designated regions. Every listed subprocessor sits in the US, but no hosting region is committed for a customer, and outside AI tools that draw on the graph process requests in their own environments.
Firms request a demo and talk to sales. A seat price, minimum subscription, usage rate, implementation fee or list of included connectors is not published.
Private equity and private credit deal teams are the market, along with the technology and AI teams inside those firms, across sourcing and screening, diligence, investment committee preparation and portfolio monitoring. Clearlake Capital, Berkshire Partners, Valesco Industries and Blue Wolf Capital are named users, and Court Square Capital appears with Metal in a PE Hub interview.
Alternatives to Metal
Other products that do the same kind of work as Metal, and what each one does.
Capsa AI sells an AI operating system to private equity firms and other private capital investors. It sources and screens targets and automates diligence tasks such as reviewing NDAs, purchase agreements and covenants and building trading comparables.
ToltIQ, formerly DiligentIQ, sells AI due diligence software to private markets deal teams. It reads virtual data rooms of thousands of documents and links every finding to its source.
F2 builds an agentic platform for private credit funds and commercial banks that covers screening, underwriting and the monitoring of loans after they close.
Keye automates the financial diligence that private equity and private credit teams run on a target company. It takes raw deal files in any format and returns structured analysis a deal team can use.
Kruncher sells private market intelligence to venture capital and private equity firms, multi family offices and secondaries investors.
Datasite Diligence is Datasite’s virtual data room for investment bankers, sellers, private equity teams, and their advisors.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
Contact sales
|
Metal prices by quote after a demo with its sales team. | Data protection terms are not tied to a published package. The trust center lists every subprocessor and its location. | Setup and connector fees are not published. | Vendor Published |
Metal offers a demo. No public subscription rate, minimum, or connector and implementation fees are stated.