Skip to content

Klarefi

by klarefi.com · remote · local (npm) · no sign-in to list tools · v0.3.0 · published by com.klarefi

Complete regulated intake work and retrieve cited case files without deciding the case.

Deletes · 9Changes · 4

21 tools: 9 can delete or cancel things, 4 change things and 8 only read.

Scanned Oct 21, 2026

2 checks matched something in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http klarefi https://www.klarefi.com/mcp

Findings

Matches of fixed checks, with the text they matched.

1 warn1 info
  • The server marks "open_intake" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "open_intake" says "open": a word for a change.

    What this check looks for
  • The description mentions "credential".

    Matched text

    description: "...d MCP endpoint, include the same intake credential in each later intake tool call."

    What this check looks for

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

21 tools

Deletes or cancels 9Claude Code: deny
  • answer_fieldsAnswer structured requested facts by fact_id.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Answer structured requested facts by fact_id. Values must be strings; yes_no fields take "yes" or "no", and date fields use YYYY-MM-DD.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • answer_questionAnswer the active applicant question with free text, optionally targeting a task_id.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Answer the active applicant question with free text, optionally targeting a task_id.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • create_connectorCreate or update a Klarefi verification connector manifest.No findings

    Why it is in this group

    • The name "create_connector" says "create": a word for a change.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Create or update a Klarefi verification connector manifest. The result reports whether the configuration is active or staged for operator approval.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • create_connector_from_openapiImport a public OpenAPI JSON URL as a Klarefi verification connector.Reads text others writeNo findings

    Why it is in this group

    • The name "create_connector_from_openapi" says "create": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Import a public OpenAPI JSON URL as a Klarefi verification connector. JSON specs only (convert YAML first). The result reports whether the configuration is active or staged for operator approval.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • delete_connectorDelete a configured Klarefi connector by key or stage its deletion for operator approval.No findings

    Why it is in this group

    • The name "delete_connector" says "delete": a word for removing or ending something.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Delete a configured Klarefi connector by key or stage its deletion for operator approval.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • save_intake_draftSave applicant form draft fields without submitting.No findings

    Why it is in this group

    • The name "save_intake_draft" says "save": a word for a change.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Save applicant form draft fields without submitting. Field values must be strings or null.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • save_workflow_draftValidate and save a workflow declaration as a draft.No findings

    Why it is in this group

    • The name "save_workflow_draft" says "save": a word for a change.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Validate and save a workflow declaration as a draft. This tool cannot publish or activate the workflow.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • submit_case_reviewSubmit an auditable intake-review command.No findings

    Why it is in this group

    • The name "submit_case_review" says "submit": a word for a change.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Submit an auditable intake-review command. Allowed commands review or correct facts, update evidence, or escalate. This tool cannot complete or reopen a case.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • submit_intakeSubmit the applicant intake.No findings

    Why it is in this group

    • The name "submit_intake" says "submit": a word for a change.
    • The server marks it destructive (destructiveHint: true).
    Description, as the server wrote it
    Submit the applicant intake. Returns submitted, blockers, and the fresh intake state.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
Changes things 4Claude Code: ask
  • create_intake_sessionCreate a hosted Klarefi intake session.No findings

    Why it is in this group

    • The name "create_intake_session" says "create": a word for a change.
    Description, as the server wrote it
    Create a hosted Klarefi intake session.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
  • open_intakeRead an applicant intake from a signed URL or from session_id plus access_token.Reads text others writeCan send data out2 findings

    Why it is in this group

    • The name "open_intake" says "open": a word for a change.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    What it can reach

    • untrusted-inputThe description's first sentence says "read" and "url": it reads web pages, which anyone can write.
    • untrusted-inputThe input takes a URL (input.url), and the tool reads it: it reads web pages, which anyone can write.
    • outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.

    Findings

    • Info

      The description mentions "credential".

      Matched text

      description: "...d MCP endpoint, include the same intake credential in each later intake tool call."

      What this check looks for
    • Warn

      The server marks "open_intake" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "open_intake" says "open": a word for a change.

      What this check looks for
    Description, as the server wrote it
    Read an applicant intake from a signed URL or from session_id plus access_token. No Klarefi API key is required. On the hosted MCP endpoint, include the same intake credential in each later intake tool call.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • upload_documentUpload a requested applicant document.Can send data outNo findings

    Why it is in this group

    • The name "upload_document" says "upload": a word for a change.

    What it can reach

    • outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
    Description, as the server wrote it
    Upload a requested applicant document. Use content_base64 on the hosted MCP endpoint. Local stdio clients can use file_path. Inline content is limited to 5MB decoded; local files can be up to 25MB.

    Annotations the server declares

    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
Only reads 8Claude Code: allow, ask for 2 outbound
  • get_caseRetrieve a Klarefi case and resolved facts.No findings

    Why it is in this group

    • The name "get_case" says "get": a word for reading.
    Description, as the server wrote it
    Retrieve a Klarefi case and resolved facts.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_case_packageGet the cited case_package.v1 artifact and a signed read-only case-file URL.Reads private dataReads text others writeNo findings

    Why it is in this group

    • The name "get_case_package" says "get": a word for reading.

    What it can reach

    • private-dataThe description's first sentence says "get" and "file": it reads files.
    • untrusted-inputThe description's first sentence says "get" and "url": it reads web pages, which anyone can write.
    Description, as the server wrote it
    Get the cited case_package.v1 artifact and a signed read-only case-file URL. The package carries the facts, the calculated facts, and the offer or decline on the case with its files.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_case_workspaceInspect a case workspace with facts, cited evidence, blockers, ownership, and timeline.No findings

    Why it is in this group

    • The name "get_case_workspace" says "get": a word for reading.
    Description, as the server wrote it
    Inspect a case workspace with facts, cited evidence, blockers, ownership, and timeline. This does not mutate the case.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_intake_stateGet the current applicant intake state.Reads text others writeCan send data outNo findings

    Why it is in this group

    • The name "get_intake_state" says "get": a word for reading.

    What it can reach

    • untrusted-inputThe input takes a URL (input.url), and the tool reads it: it reads web pages, which anyone can write.
    • outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.

    Claude Code: ask, because it is tagged outbound.

    Description, as the server wrote it
    Get the current applicant intake state. Pass url, or session_id with access_token, on the hosted MCP endpoint.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_review_commandPoll the status and result of an intake-review command.Reads text others writeNo findings

    Why it is in this group

    • The name "get_review_command" says "get": a word for reading.

    What it can reach

    • untrusted-inputThe name "get_review_command" says "get" and "review": it reads comments, which other people can write.
    • untrusted-inputThe description's first sentence says "review": it reads comments, which other people can write.
    Description, as the server wrote it
    Poll the status and result of an intake-review command.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • list_operator_queueList the bounded Klarefi operator queue for this API key's environment.No findings

    Why it is in this group

    • The name "list_operator_queue" says "list": a word for reading.
    Description, as the server wrote it
    List the bounded Klarefi operator queue for this API key's environment. Requires cases:review.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • list_workflowsList versioned Form · Fact · Verify workflow declarations.No findings

    Why it is in this group

    • The name "list_workflows" says "list": a word for reading.
    Description, as the server wrote it
    List versioned Form · Fact · Verify workflow declarations.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • wait_for_next_actionPoll until the intake leaves processing or the timeout expires.Reads text others writeCan send data outNo findings

    Why it is in this group

    • The name "wait_for_next_action" says "wait": a word for reading.

    What it can reach

    • untrusted-inputThe input takes a URL (input.url), and the tool reads it: it reads web pages, which anyone can write.
    • outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.

    Claude Code: ask, because it is tagged outbound.

    Description, as the server wrote it
    Poll until the intake leaves processing or the timeout expires. If timed_out is true, call wait_for_next_action again.

    Annotations the server declares

    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Exfil surface

Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.

30 paths in this server alone from private data, through text someone else wrote, to a way out. Denying 1 private-data tool breaks every one.

Reads private data1
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write6
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out5
Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
deny to break every path
{
  "permissions": {
    "deny": [
      "mcp__klarefi__get_case_package"
    ]
  }
}

Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.

Remote posture

What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.

Looked at Oct 22, 2026, 14:34 UTC, unchanged since Oct 22, 2026. No credentials were sent and no tool was called. What is looked at.

Reachable
yes
HTTP status of an anonymous initialize
404
tools/list served without credentials
no
Tools that change things, listed without credentials
0
Protocol version
none
WWW-Authenticate on 401
not observed
resource_metadata in WWW-Authenticate
none
RFC 9728 protected resource metadata
none
Authorization servers
none
Legacy HTTP+SSE transport
no
Dynamic client registration
not observed
Client ID metadata documents
not observed

The probe stopped: initialize answered HTTP 404.

Install

Pinned to version 0.3.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http klarefi https://www.klarefi.com/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__klarefi__get_case",
      "mcp__klarefi__list_operator_queue",
      "mcp__klarefi__get_case_workspace",
      "mcp__klarefi__get_case_package",
      "mcp__klarefi__get_review_command",
      "mcp__klarefi__list_workflows"
    ],
    "ask": [
      "mcp__klarefi__create_intake_session",
      "mcp__klarefi__open_intake",
      "mcp__klarefi__get_intake_state",
      "mcp__klarefi__upload_document",
      "mcp__klarefi__give_intake_consent",
      "mcp__klarefi__wait_for_next_action"
    ],
    "deny": [
      "mcp__klarefi__create_connector_from_openapi",
      "mcp__klarefi__create_connector",
      "mcp__klarefi__delete_connector",
      "mcp__klarefi__submit_case_review",
      "mcp__klarefi__save_workflow_draft",
      "mcp__klarefi__answer_question",
      "mcp__klarefi__answer_fields",
      "mcp__klarefi__submit_intake",
      "mcp__klarefi__save_intake_draft"
    ]
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as klarefi. Review it before you use it.

Capability tags behind 8 rules
  • mcp__klarefi__create_connector_from_openapi untrusted-input
  • mcp__klarefi__get_case_package private-data, untrusted-input
  • mcp__klarefi__get_review_command untrusted-input
  • mcp__klarefi__open_intake untrusted-input, outbound
  • mcp__klarefi__get_intake_state untrusted-input, outbound
  • mcp__klarefi__upload_document outbound
  • mcp__klarefi__give_intake_consent outbound
  • mcp__klarefi__wait_for_next_action untrusted-input, outbound