DaubertDaubertRequest access
Model Context Protocol

Bring your own agent
to the case file.

Daubert speaks MCP. Connect Claude Desktop, claude.ai, Cowork, or Claude Code and your agent reads the case, pulls chain data, and writes productions back. Same roles, same audit trail, same case file.

Request accessConnect it in two minutes
tools
19tools
workflow prompts
7workflow prompts
chains live
7chains live
daubert · connected
›list_cases
3 cases · Ronin bridge, Doe v. Roe, estate recovery
›blockchain_fetch_history { address: "0x9f2c…41ab", chain: "ethereum" }
1,284 transfers · 2023-04-11 to 2024-02-08
›import_transactions { traceId: "b17e…", transactions: [ … ] }
added 1,284 · 96 new nodes · logged to the case
›create_production { type: "declaration", formatId: "federal-1746" }
declaration skeleton created
What it unlocks

Three jobs, start to finish, without leaving your agent.

The tools are small on purpose. Chained, they cover the work an analyst actually does on a matter.

01 · Trace

Pull a wallet and put it on the graph.

Your agent finds the case, fetches history from the chain with Daubert's API keys, and writes the transactions into a trace. Unknown addresses become nodes automatically.

  1. 01list_cases
  2. 02get_investigation
  3. 03blockchain_fetch_history
  4. 04import_transactions
02 · Declare

Draft a declaration on the record.

It reads the declarant profile and your boilerplate library, then builds the declaration section by section. Five jurisdiction formats, with the oath and caption rendered for you.

  1. 01get_declarants
  2. 02get_declaration_library
  3. 03create_production
  4. 04update_production
03 · Redline

Review a draft against the evidence.

Point it at a draft in the data room. It snapshots the text, checks each claim against the graph and the case record, and proposes tracked changes with anchors quoted verbatim.

  1. 01list_data_room_files
  2. 02read_data_room_file
  3. 03create_production
  4. 04update_production
The tool set

Nineteen tools. Every one gated by your role.

An agent connected to Daubert sees exactly what you see, and can change exactly what you can change. The gate is checked on every single call, not once at connect time.

viewer on the caseeditor on the caseorganizationany connected agent
Navigate

Find the matter.

Where an agent starts. Nothing here leaves your organization.

  • list_casesorganization

    Every case you can see in your organization, with your role on each.

  • get_caseviewer on the case

    One case by id, including its investigations.

  • list_investigationsviewer on the case

    The investigations sitting under a case.

Read

Read the case.

The graph, the productions, the entity catalog, and the data room.

  • get_case_dataviewer on the case

    The whole case in one call: investigations with trace counts, productions by name and type, and the data-room manifest.

  • get_investigationviewer on the case

    The graph itself. Nodes, edges denormalized with from and to addresses, groups, and bundles. Filter to an address or a token.

  • read_productionviewer on the case

    Read one production in full, or list every production on a case, optionally by type.

  • query_labeled_entitiesorganization

    Look up exchanges, contracts, and known wallets by address, or search by name and category.

  • list_data_room_filesviewer on the case

    The data-room manifest: id, name, type, size, and folder path, up to 500 files.

  • read_data_room_fileviewer on the case

    Read a file. docx, pdf, xlsx, csv, and txt come back as extracted text; images as an image block. PDFs up to 32 MB, other types up to 5 MB.

  • get_skillany connected agent

    Pull a workflow document by name when the agent needs the procedure mid-task.

Library

Reuse what you have written.

Organization-wide, so the agent does not re-invent your standard language.

  • get_declarantsorganization

    Saved expert witness profiles: title, firm, qualifications, prior testimony, CV exhibit, rate, and disclosures.

  • get_declaration_libraryorganization

    Reusable boilerplate blocks. Technical chain primers and authentication language.

Blockchain

Pull from the chain.

Ethereum, Polygon, Arbitrum, Base, TRON, Bitcoin, and Solana, on Daubert's server-side API keys. No keys of your own to manage.

  • blockchain_fetch_historyany connected agent

    Native transactions and token transfers for an address, merged and sorted newest first. Narrow by date range, cap the fetch.

  • blockchain_get_transactionany connected agent

    One transaction by hash: from, to, amount, token transfers, and error status.

  • blockchain_get_address_infoany connected agent

    Balance, whether the address is a wallet or a contract, and a known label if there is one.

Write

Put work back in the case.

Editor role required. Every write lands in the audit log with the session that made it.

  • create_investigationeditor on the case

    Open a new investigation under a case.

  • import_transactionseditor on the case

    Write transactions into a trace. Unknown addresses become nodes, and rows are deduplicated by hash, sender, and recipient.

  • create_productioneditor on the case

    Start a report, chart, chronology, declaration, or redline. Declarations seed the jurisdiction's oath and caption; redlines snapshot the draft they review.

  • update_productioneditor on the case

    Rename it, replace its payload, or apply atomic ops. Chronology, chart, declaration, and redline ops each have a prompt that documents them.

Prompts

The procedure ships with the tools.

Tool names alone do not teach an agent how a declaration is structured or what a valid redline anchor looks like. Daubert registers seven workflow documents as MCP prompts, so the method comes down the wire with the connection. Your agent can also pull any of them mid-task with get_skill.

This is the part that makes a general agent useful on a matter instead of merely connected to one.

  • daubert-overview

    How Daubert is organized, the full tool surface, the fetch to import workflow, and the access model. The orientation an agent reads first.

  • graph-mutations

    Adding, editing, and deleting nodes, edges, and groups inside a trace.

  • productions

    Building reports, charts, and chronologies for a case.

  • declarations

    Drafting court-ready declarations across five formats: CA declaration, NY affirmation, Federal § 1746, TX unsworn declaration, FL § 92.525.

  • redlining

    Reviewing a draft someone else wrote against the case record, and proposing tracked changes anchored to the text.

  • etherscan-apis

    Etherscan V2 reference for EVM chain queries.

  • tronscan-apis

    Tronscan and TronGrid reference for TRON queries.

Setup

Paste one URL. Sign in. Done.

Daubert supports dynamic client registration, so there is no client ID, no secret, and no config file to hand-edit. Copy your MCP server URL from Account · Connected Agents inside Daubert, then follow the track for your client.

Claude Desktop · claude.ai · Cowork
  1. 1Open Settings → Customize → Connectors.
  2. 2Scroll past the partner directory and click "+" to add a custom connector.
  3. 3Name it "Daubert" and paste your MCP server URL.
  4. 4Leave Advanced settings empty. No client ID is needed.
  5. 5Click Add. Claude opens a browser tab where you sign in and approve access.

On a Team or Enterprise plan, your workspace admin registers Daubert once under Organization settings · Connectors. Everyone else then sees a Connect button on that entry.

Claude Code
  1. 1Run the command below in your terminal.
  2. 2Run /mcp inside Claude Code and pick "daubert" to sign in through your browser.
claude mcp add --transport http daubert <your-mcp-url>

The same OAuth flow backs both tracks. A session is bound to one organization, and you can revoke it from Account · Connected Agents at any time.

Access and audit

Connecting an agent does not widen the door.

A connected agent is a session, not an account. It is bound to one organization at consent time and cannot reach another one, whatever a prompt tells it to try. Case ids from outside your organization come back as not found.

Membership is re-read from the database on every call, not cached for the life of the token. Remove someone from the organization and their agent stops working on the next request, mid-session. Guests cannot connect at all.

Reads are gated at viewer on the case. Writes are gated at editor, and each one writes an audit row naming the session, the action, and the target. The case file records what the agent did, the same way it records what you did.

Organization-bound

One session, one organization, fixed at sign-in and re-checked per call.

Role-gated

Viewer to read, editor to write. The same roles that govern the app.

Rate-limited

60 calls per minute per session, plus an IP limit before authentication runs.

Revocable

See every connected agent in your account, with its client and last use. Revoke one instantly.

Get early access

Point your agent at a real case.

We’re onboarding a small group of design partners through the preview. Tell us the matter you’d run through it.

  • Free during preview. No card, no quota.
  • Works with the Claude client you already use.
  • You keep ownership of your data.

Already have an account? Open Account · Connected Agents in Daubert, copy your MCP server URL, and follow the setup steps.

We read every one. Replies come from a real person, not a sequence.