Skip to content

AgentDrive MCP server

sh.agentdrive/agentdriveagentdrive.shstreamable-httpv1.4.1Scanned

Sync coding-agent sessions and resume them in any other agent: Claude Code, Codex, Cursor, more.

22 tools: 1 can delete or cancel things, 1 do something that can't be undone, 10 change things and 10 only read.

Tools: 1 delete, 1 irreversible, 10 write, 10 read.
  • Reads private data · 10
  • Reads text others write · 2
  • Can send data out · 2
  • No findings

Sign-in required, OAuth metadata published. Remote posture

With Anlyon

1 tool here can delete or cancel things and 1 does something that can't be undone. Govern them.

  • Anlyon lists the tools and adopts none.
  • Each tool you adopt gets approvals and receipts.
  • Your agent never holds the key.
Govern this server →

A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.

What the AgentDrive MCP server can do

The AgentDrive MCP server is published by agentdrive.sh, a domain the MCP Registry verified. It has 22 tools. It can delete or cancel things (1 tool: drive_delete), do things that cannot be undone, such as sending, paying or publishing (drive_invite) and create or change data (10 tools: drive_write, drive_share, drive_comment, drive_roles and 6 more). 10 tools only read.

10 tools read private data (drive_read, drive_list, drive_feedback and 7 more); 2 tools bring in text other people can write (drive_feedback and drive_activity); 2 tools can send data out of your system (drive_invite and drive_sandbox_clone). On its own it has all three parts of a prompt-injection exfiltration path: private data, text others write, and a way out.

None of the fixed checks matched anything in its tool definitions. That is not a safety guarantee.

Its remote endpoint asks for sign-in and publishes OAuth metadata, and its tools changed once since Anlyon Radar first saw them.

Last scanned Nov 4, 2026, version 1.4.1.

Tools

Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.

22 tools

Deletes or cancels 1Claude Code: deny
  • drive_deleteDelete a file from a workspace.No findings

    Why it is in this group

    • The name "drive_delete" says "delete": a word for removing or ending something.
    Description, as the server wrote it
    Delete a file from a workspace.
Can't be undone 1Claude Code: ask
  • drive_inviteInvite a human to your organization by email.Can send data outNo findings

    Why it is in this group

    • The name "drive_invite" says "invite": a word for sending, and what is sent cannot be taken back.

    What it can reach

    • outboundThe name "drive_invite" says "invite": it sends or publishes something to other people.
    Description, as the server wrote it
    Invite a human to your organization by email. The invite stays pending until they sign in and accept it.
Changes things 10Claude Code: ask
  • drive_commentAdd a comment to a file.No findings

    Why it is in this group

    • The name "drive_comment" says "comment": a word for a change.
    Description, as the server wrote it
    Add a comment to a file.
  • drive_contextAuto-pull the most relevant artifacts from prior agent sessions for the task you are about to do.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.
    Description, as the server wrote it
    Auto-pull the most relevant artifacts from prior agent sessions for the task you are about to do. Combines semantic relevance with recency. Call this at the START of any non-trivial task to ground the agent in prior context — replaces manual copy-paste of artifacts between sessions.
  • drive_recallRecall persistent facts and preferences your agents have written.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.
    Description, as the server wrote it
    Recall persistent facts and preferences your agents have written. In v1 this returns semantic search results across artifacts; v1.5 will narrow to auto-extracted facts (preferences, decisions, key entities).
  • drive_resumeLoad a compact handoff brief (goal, files touched, recent turns) from a prior session, by default the latest one from a different harness, so you can continue that work.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.
    Description, as the server wrote it
    Load a compact handoff brief (goal, files touched, recent turns) from a prior session, by default the latest one from a different harness, so you can continue that work.
  • drive_rolesManage per-workspace role delegation.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.
    Description, as the server wrote it
    Manage per-workspace role delegation. Assign, list, or remove access for specific emails (view/comment/edit).
  • drive_sandbox_clone[EXPERIMENTAL] Clone a git repo into a per-workspace Linux sandbox.Can send data outNo findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.

    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
    [EXPERIMENTAL] Clone a git repo into a per-workspace Linux sandbox. Requires Cloudflare Containers on the AgentDrive worker; gracefully degrades to a 503 with "sandbox not configured" otherwise.
  • drive_sandbox_commit[EXPERIMENTAL] Stage + commit the current sandbox worktree.No findings

    Why it is in this group

    • The name "drive_sandbox_commit" says "commit": a word for a change.
    Description, as the server wrote it
    [EXPERIMENTAL] Stage + commit the current sandbox worktree. Surfaces as an AgentDrive artifact at commits/<sha>.json. Requires Cloudflare Containers; degrades to 503 otherwise.
  • drive_sandbox_exec[EXPERIMENTAL] Run a shell command inside the per-workspace Linux sandbox.No findings

    Why it is in this group

    • The name "drive_sandbox_exec" says "exec": a word for a change.
    • The input takes a statement or command to run (input.cmd), which can change anything it can reach.
    Description, as the server wrote it
    [EXPERIMENTAL] Run a shell command inside the per-workspace Linux sandbox. Captures stdout/stderr (64 KB cap), exit code, duration. Requires Cloudflare Containers; degrades to 503 otherwise.
  • drive_shareGenerate a share link for a workspace or specific file.No findings

    Why it is in this group

    • The description's first verb is "generate": a word for a change.
    Description, as the server wrote it
    Generate a share link for a workspace or specific file. Optionally email-gate the link so only a specific person can access it. Humans open this link to view, comment, or edit files.
  • drive_writeWrite a markdown file to a workspace.No findings

    Why it is in this group

    • The name "drive_write" says "write": a word for a change.
    Description, as the server wrote it
    Write a markdown file to a workspace. Creates or updates the file with automatic versioning.
Only reads 10Show 10 read-only tools
  • drive_activityGet recent activity in a workspace (file writes, comments, shares).Reads private dataReads text others writeNo findings

    Why it is in this group

    • The description's first verb is "get": a word for reading.

    What it can reach

    • private-dataThe name "drive_activity" says "drive": it reads files.
    • private-dataThe description's first sentence says "get" and "file": it reads files.
    • untrusted-inputThe description's first sentence says "get" and "comment": it reads comments, which other people can write.
    Description, as the server wrote it
    Get recent activity in a workspace (file writes, comments, shares).
  • drive_feedbackRead human comments on a file.Reads private dataReads text others writeNo findings

    Why it is in this group

    • The description's first verb is "read": a word for reading.

    What it can reach

    • private-dataThe name "drive_feedback" says "drive": it reads files.
    • private-dataThe description's first sentence says "read" and "file": it reads files.
    • private-dataThe input names a file (input.file_id), and the tool reads: it reads files.
    • untrusted-inputThe description's first sentence says "read" and "comment": it reads comments, which other people can write.
    Description, as the server wrote it
    Read human comments on a file. Use this to check if humans have left feedback.
  • drive_listList files in a workspace, optionally filtered by path prefix.Reads private dataNo findings

    Why it is in this group

    • The name "drive_list" says "list": a word for reading.

    What it can reach

    • private-dataThe name "drive_list" says "list" and "drive": it reads files.
    • private-dataThe description's first sentence says "list" and "file": it reads files.
    Description, as the server wrote it
    List files in a workspace, optionally filtered by path prefix.
  • drive_log_runLog an agent run to AgentDrive's provenance store.Reads private dataNo findings

    Why it is in this group

    • The name "drive_log_run" says "log": a word for reading.

    What it can reach

    • private-dataThe name "drive_log_run" says "log" and "drive": it reads files.
    • private-dataThe description's first sentence says "log" and "drive": it reads files.
    Description, as the server wrote it
    Log an agent run to AgentDrive's provenance store. Captures task, status, referenced/produced artifacts, and cross-platform identity (which agent on which tool). Used for retrospective context retrieval and team-wide audit.
  • drive_orgView your organization details: agents, members, plan, and usage.Reads private dataNo findings

    Why it is in this group

    • The description's first verb is "view": a word for reading.

    What it can reach

    • private-dataThe name "drive_org" says "drive": it reads files.
    Description, as the server wrote it
    View your organization details: agents, members, plan, and usage.
  • drive_readRead a file's content from a workspace.Reads private dataNo findings

    Why it is in this group

    • The name "drive_read" says "read": a word for reading.

    What it can reach

    • private-dataThe name "drive_read" says "read" and "drive": it reads files.
    • private-dataThe description's first sentence says "read" and "file": it reads files.
    • private-dataThe input names a file (input.file_id), and the tool reads: it reads files.
    Description, as the server wrote it
    Read a file's content from a workspace.
  • drive_sandbox_read[EXPERIMENTAL] Read a file from the per-workspace sandbox filesystem (build artifacts, test output, logs).Reads private dataNo findings

    Why it is in this group

    • The name "drive_sandbox_read" says "read": a word for reading.

    What it can reach

    • private-dataThe name "drive_sandbox_read" says "read" and "drive": it reads files.
    • private-dataThe description's first sentence says "file": it reads files.
    Description, as the server wrote it
    [EXPERIMENTAL] Read a file from the per-workspace sandbox filesystem (build artifacts, test output, logs). Requires Cloudflare Containers; degrades to 503 otherwise.
  • drive_sessionsList recent agent sessions for this project across harnesses (Claude Code, Codex, OpenCode, Gemini CLI, Cursor, Antigravity), newest first.Reads private dataNo findings

    Why it is in this group

    • The description's first verb is "list": a word for reading.

    What it can reach

    • private-dataThe name "drive_sessions" says "drive": it reads files.
    Description, as the server wrote it
    List recent agent sessions for this project across harnesses (Claude Code, Codex, OpenCode, Gemini CLI, Cursor, Antigravity), newest first. Use to find prior work to continue.
  • drive_versionsView version history for a file in a workspace.Reads private dataNo findings

    Why it is in this group

    • The description's first verb is "view": a word for reading.

    What it can reach

    • private-dataThe name "drive_versions" says "drive": it reads files.
    • private-dataThe description's first sentence says "view" and "file": it reads files.
    • private-dataThe input names a file (input.file_id), and the tool reads: it reads files.
    Description, as the server wrote it
    View version history for a file in a workspace.

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.

40 paths in this server alone from private data, through text someone else wrote, to a way out. Denying 2 outbound tools breaks every one.

Reads private data10
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write2
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out2
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__agentdrive__drive_invite",
      "mcp__agentdrive__drive_sandbox_clone"
    ]
  }
}

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.

Sign-in required, OAuth metadata published

Looked at Nov 20, 2026, 19:18 UTC, unchanged since Nov 13, 2026. No credentials were sent and no tool was called. What is looked at.

Reachable
yes
HTTP status of an anonymous initialize
401
tools/list served without credentials
no
Tools that change things, listed without credentials
0
Protocol version
not observed
WWW-Authenticate on 401
yes
resource_metadata in WWW-Authenticate
https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
RFC 9728 protected resource metadata
https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
Authorization servers
https://agentdrive.sh
Legacy HTTP+SSE transport
no
Dynamic client registration
yes
Client ID metadata documents
yes

Posture changes

  1. Nov 13, 2026
    • HTTP status of an anonymous initialize: 200 to 401
    • tools/list served without credentials: yes to no
    • Tools that change things, listed without credentials: 12 to 0
    • Protocol version: 2025-06-18 to none
    • WWW-Authenticate on 401: not observed to yes
    • resource_metadata in WWW-Authenticate: none to https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
    • RFC 9728 protected resource metadata: none to https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
    • Authorization servers: none to https://agentdrive.sh
    • Dynamic client registration: not observed to yes
    • Client ID metadata documents: not observed to yes

Findings

Matches of fixed checks, with the text they matched.

No check matched anything in these tool definitions. That is not a safety guarantee.

History

Changes to its tool definitions and to its remote endpoint, newest first.

  1. PostureRemote posture changed. HTTP status of an anonymous initialize: 200 -> 401; tools/list served without credentials: yes -> no; Tools that change things, listed without credentials: 12 -> 0 (and 7 more).View diff
    • HTTP status of an anonymous initialize: 200 to 401
    • tools/list served without credentials: yes to no
    • Tools that change things, listed without credentials: 12 to 0
    • Protocol version: 2025-06-18 to none
    • WWW-Authenticate on 401: not observed to yes
    • resource_metadata in WWW-Authenticate: none to https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
    • RFC 9728 protected resource metadata: none to https://mcp.agentdrive.sh/.well-known/oauth-protected-resource/mcp
    • Authorization servers: none to https://agentdrive.sh
    • Dynamic client registration: not observed to yes
    • Client ID metadata documents: not observed to yes
  2. Tools3 changed.~3 changedView diff

Install

Pinned to version 1.4.1. Values in angle brackets are yours to fill in.

claude mcp add --transport http agentdrive https://mcp.agentdrive.sh/mcp --header 'Authorization: <your-authorization>'
claude mcp add --transport http agentdrive https://mcp.agentdrive.sh/mcp --header 'Authorization: <your-authorization>'

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__agentdrive__drive_read",
      "mcp__agentdrive__drive_list",
      "mcp__agentdrive__drive_feedback",
      "mcp__agentdrive__drive_activity",
      "mcp__agentdrive__drive_org",
      "mcp__agentdrive__drive_versions",
      "mcp__agentdrive__drive_search",
      "mcp__agentdrive__drive_log_run",
      "mcp__agentdrive__drive_sessions",
      "mcp__agentdrive__drive_sandbox_read"
    ],
    "ask": [
      "mcp__agentdrive__drive_write",
      "mcp__agentdrive__drive_share",
      "mcp__agentdrive__drive_comment",
      "mcp__agentdrive__drive_invite",
      "mcp__agentdrive__drive_roles",
      "mcp__agentdrive__drive_recall",
      "mcp__agentdrive__drive_context",
      "mcp__agentdrive__drive_resume",
      "mcp__agentdrive__drive_sandbox_clone",
      "mcp__agentdrive__drive_sandbox_exec",
      "mcp__agentdrive__drive_sandbox_commit"
    ],
    "deny": [
      "mcp__agentdrive__drive_delete"
    ]
  }
}

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 agentdrive. Review it before you use it.

Capability tags behind 12 rules
  • mcp__agentdrive__drive_read private-data
  • mcp__agentdrive__drive_list private-data
  • mcp__agentdrive__drive_feedback private-data, untrusted-input
  • mcp__agentdrive__drive_activity private-data, untrusted-input
  • mcp__agentdrive__drive_invite outbound
  • mcp__agentdrive__drive_org private-data
  • mcp__agentdrive__drive_versions private-data
  • mcp__agentdrive__drive_search private-data
  • mcp__agentdrive__drive_log_run private-data
  • mcp__agentdrive__drive_sessions private-data
  • mcp__agentdrive__drive_sandbox_clone outbound
  • mcp__agentdrive__drive_sandbox_read private-data

Pin

Keep the decision: fail CI when the tools change, follow every change, or show a badge.

Pin in CI

Fail a build when the tools you reviewed are not the tools the server serves today.

curl -s "https://radar.anlyon.com/api/v1/servers/sh.agentdrive/agentdrive/verify?hash=5b15770aa627bc54c123b5aa99937514db8d0d42ba28b1fe0f28799a8677a8a1" | jq -e '.changed == false'

Pinned hash 5b15770aa627, from Nov 4, 2026, 14:17 UTC.

Watch

Get told when its tools change: a new tool, a new description, a new class.

https://radar.anlyon.com/servers/sh.agentdrive/agentdrive/changes.xml

An Atom feed. Works in any feed reader or Slack's RSS app.

Badge

Anlyon Radar badge for sh.agentdrive/agentdrive
[![Anlyon Radar](https://radar.anlyon.com/badge/sh.agentdrive/agentdrive.svg)](https://radar.anlyon.com/servers/sh.agentdrive/agentdrive)

Source: MCP Registry · repository. Synced Nov 21, 2026, 10:18 UTC.

AgentDrive MCP server: questions and answers

Answered from the scan above, not from the publisher's description.

Is the AgentDrive MCP server safe?

No scan can promise that. What its definitions show: 22 tools, 12 of which can change, send or delete something, and 0 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.

Can the AgentDrive MCP server delete data?

Yes. 1 tool can delete, cancel or revoke something: drive_delete.

Can the AgentDrive MCP server do things that cannot be undone?

Yes. 1 tool can send, pay, publish or merge: drive_invite.

What tools does the AgentDrive MCP server have?

22 tools: 1 destructive, 1 irreversible, 10 write, 10 read. drive_delete, drive_invite, drive_write, drive_share, drive_comment, drive_roles, drive_recall, drive_context and 14 more.

Does the AgentDrive MCP server need sign-in or OAuth?

Yes. Its remote endpoint answered 401 without credentials, and it publishes OAuth protected resource metadata (RFC 9728); its authorization server supports dynamic client registration.

Does the AgentDrive MCP server send data to outside services?

2 tools can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): drive_invite and drive_sandbox_clone.

Has the AgentDrive MCP server changed its tools?

Yes, once since Anlyon Radar started tracking it, most recently on Nov 4, 2026. Every change is on this page and in its Atom feed.