AgentDrive

agentdrive-sh/agentdrive-mcp

AgentDrive gives agents persistent memory and artifact storage that survives across tools and sessions. It exposes 20 MCP tools split across files (write, read, list, versions), semantic search and recall, provenance logging, and workspace collaboration. Write a markdown spec in Claude Code, then have Cursor's agent pick it up with drive_context or drive_search in the next session. Run logs track which agent on which platform touched what, so handoffs between Claude Code, Codex, Cursor, and LangGraph maintain attribution. Free tier includes 1 GB storage and 10k embeddings. Available as a hosted streamable HTTP endpoint or local stdio via npx.

GitHub stars: 2
tools: 22
auth: Secret needed
transport: STDIO, HTTP
env vars: 2
last updated: Nov 10, 2026

View on GitHub

Install

From registry metadata

$ npx add-mcp https://mcp.agentdrive.sh/mcp --name agentdrive --header 'Authorization: Bearer ${API_TOKEN}'

Run in your terminal. add-mcp asks which agents to add it to, then asks for each value it needs. Add -g to install it for every project.

Review the command, arguments and environment values before installing. MCP servers run with your local permissions.

Other ways

Claude Code

$ claude mcp add --transport http agentdrive https://mcp.agentdrive.sh/mcp --header 'Authorization: YOUR_AUTHORIZATION'

Run in your terminal. Add --scope user to make it available in every project.

Tools

Verified live · 22 tools

Verified live against the running server on Nov 6, 2026.

drive_writeWrite a markdown file to a workspace. Creates or updates the file with automatic versioning.4 params

Write a markdown file to a workspace. Creates or updates the file with automatic versioning.

Parameters* required

content*string

File content (markdown)

messagestring

Optional version message

path*string

File path (e.g. "notes/meeting.md")

workspace_id*string

Workspace ID to write to

drive_readRead a file's content from a workspace.2 params

Read a file's content from a workspace.

Parameters* required

file_id*string

File ID to read

workspace_id*string

Workspace ID

drive_listList files in a workspace, optionally filtered by path prefix.2 params

List files in a workspace, optionally filtered by path prefix.

Parameters* required

prefixstring

Path prefix filter (e.g. "notes/")

workspace_id*string

Workspace ID

drive_shareGenerate 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.6 params

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.

Parameters* required

emailstring

Email-gate: only this email can access the link (requires email verification). Leave empty for public access.

expires_innumber

Expiry in seconds (optional)

file_idstring

Scope to a specific file (optional)

max_usesnumber

Maximum number of uses (optional)

permission*string

Permission level for the share linkone of view · comment · edit

workspace_id*string

Workspace ID

drive_commentAdd a comment to a file.5 params

Add a comment to a file.

Parameters* required

body*string

Comment text

file_id*string

File ID to comment on

line_endnumber

End line number (optional)

line_startnumber

Start line number (optional)

workspace_id*string

Workspace ID

drive_feedbackRead human comments on a file. Use this to check if humans have left feedback.2 params

Read human comments on a file. Use this to check if humans have left feedback.

Parameters* required

file_id*string

File ID to read comments from

workspace_id*string

Workspace ID

drive_activityGet recent activity in a workspace (file writes, comments, shares).1 params

Get recent activity in a workspace (file writes, comments, shares).

Parameters* required

workspace_id*string

Workspace ID

drive_inviteInvite a human to your organization by email. The invite stays pending until they sign in and accept it.2 params

Invite a human to your organization by email. The invite stays pending until they sign in and accept it.

Parameters* required

email*string

Email address of the human to invite to your organization

rolestring

Role in the organization (default: member)one of owner · admin · memberdefault: member

drive_rolesManage per-workspace role delegation. Assign, list, or remove access for specific emails (view/comment/edit).4 params

Manage per-workspace role delegation. Assign, list, or remove access for specific emails (view/comment/edit).

Parameters* required

action*string

Action: assign a role, list all roles, or remove a roleone of assign · list · remove

emailstring

Email address (required for assign/remove)

permissionstring

Permission level (required for assign)one of view · comment · edit

workspace_id*string

Workspace ID

drive_orgView your organization details: agents, members, plan, and usage.

View your organization details: agents, members, plan, and usage.

No parameters. Call it with no arguments.

drive_deleteDelete a file from a workspace.2 params

Delete a file from a workspace.

Parameters* required

file_id*string

File ID to delete

workspace_id*string

Workspace ID

drive_versionsView version history for a file in a workspace.2 params

View version history for a file in a workspace.

Parameters* required

file_id*string

File ID

workspace_id*string

Workspace ID

drive_searchSemantic search across your org's artifacts in AgentDrive. Returns the top-N most relevant files. Use this BEFORE writing similar artifacts to avoid duplication and surface prior context.4 params

Semantic search across your org's artifacts in AgentDrive. Returns the top-N most relevant files. Use this BEFORE writing similar artifacts to avoid duplication and surface prior context.

Parameters* required

filterobject

query*string

Search query

top_kinteger

Max hits to return (default 10, max 50)

workspace_idstring

Scope to a single workspace; omit for org-wide search

drive_recallRecall 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).2 params

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).

Parameters* required

query*string

Recall query — what fact, preference, or decision are you trying to remember?

top_kinteger

Max results (default 5, max 20)

drive_log_runLog 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.7 params

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.

Parameters* required

artifacts_producedarray

Paths of artifacts the run wrote

artifacts_referencedarray

Paths of artifacts the run read

duration_msinteger

Wall-clock duration in milliseconds

metadataobject

Free-form key/value metadata

run_idstring

Existing run ID to update; omit to let server allocate

statusstring

Run outcome (default success)one of success · failure · partial · in_progress

task*string

What the agent did or is doing (free text)

drive_contextAuto-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 ses...3 params

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 ses...

Parameters* required

limitinteger

Max artifacts to return (default 5, max 20)

task*string

What you are about to do — used to rank prior artifacts by relevance

workspace_idstring

Scope to a single workspace; omit for org-wide context

drive_sessionsList 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.3 params

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.

Parameters* required

limitinteger

Max sessions to return (default 10, max 50)

platformstring

Only sessions from this harness (claude-code, codex, opencode, gemini-cli, cursor, antigravity)

projectstring

Project key; defaults to the project of the current working directory

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.3 params

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.

Parameters* required

max_charsinteger

Handoff size budget in characters (default 12000)

platformstring

When session_id is omitted, only consider sessions from this harness

session_idstring

Session run id from drive_sessions (e.g. codex:abc123); omit to pick the latest session from another harness

drive_sandbox_clone[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.4 params

[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.

Parameters* required

branchstring

Branch to check out (default: repo default)

depthinteger

Shallow clone depth (default: full clone)

url*string

Git repository URL to clone

workspace_id*string

Workspace to clone into. The sandbox persists across MCP sessions per workspace.

drive_sandbox_exec[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.4 params

[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.

Parameters* required

cmd*string

Shell command to run inside the per-workspace Linux sandbox

cwdstring

Working directory (default: workspace root)

timeout_msinteger

Max runtime in ms (default 30000, hard cap 60000)

workspace_id*string

Workspace whose sandbox to run in

drive_sandbox_read[EXPERIMENTAL] Read a file from the per-workspace sandbox filesystem (build artifacts, test output, logs). Requires Cloudflare Containers; degrades to 503 otherwise.2 params

[EXPERIMENTAL] Read a file from the per-workspace sandbox filesystem (build artifacts, test output, logs). Requires Cloudflare Containers; degrades to 503 otherwise.

Parameters* required

path*string

Path inside the sandbox (e.g., src/index.ts or /tmp/build.log)

workspace_id*string

Workspace whose sandbox to read from

drive_sandbox_commit[EXPERIMENTAL] Stage + commit the current sandbox worktree. Surfaces as an AgentDrive artifact at commits/<sha>.json. Requires Cloudflare Containers; degrades to 503 otherwise.4 params

[EXPERIMENTAL] Stage + commit the current sandbox worktree. Surfaces as an AgentDrive artifact at commits/<sha>.json. Requires Cloudflare Containers; degrades to 503 otherwise.

Parameters* required

author_emailstring

Override git author email (default: agent identity)

author_namestring

Override git author name (default: agent identity)

message*string

Commit message

workspace_id*string

Workspace whose sandbox to commit

Configuration

AGENTDRIVE_API_KEYsecret

AgentDrive API key. Optional after `npx @adrive/cli auth login`.

AGENTDRIVE_AGENT_PLATFORM

Harness id recorded as provenance, e.g. claude-code or codex.

Readme

AgentDrive MCP Server

The official Model Context Protocol server for AgentDrive: shared workspace for coding agents. Capture sessions from Claude Code, Codex, OpenCode, Cursor, Gemini CLI and Antigravity, resume them in any of the others, and search everything they wrote. Provenance on every artifact; portable storage on R2, S3 or your own bucket.

Hit a usage limit in one agent? Call drive_resume in another and it loads the handoff: the goal, the files touched and the latest turns. 22 MCP tools (18 stable + 4 experimental sandbox), over stdio or hosted. Free plan, no credit card.

Website: https://agentdrive.sh
Docs: https://agentdrive.sh/docs
npm: @adrive/mcp-server
Hosted endpoint: https://mcp.agentdrive.sh/mcp

Claude Code plugin

This repo is also the Claude Code plugin marketplace. The plugin captures every session in the background and resumes it from Codex, OpenCode, Gemini CLI, Antigravity, or Cursor:

/plugin install agentdrive --marketplace agentdrive-sh/agentdrive-mcp

View the full README on GitHub

Details

Repository: github.com/agentdrive-sh/agentdrive-mcp
Website: agentdrive.sh
Category: AI & LLM Tools
Package: @adrive/mcp-server
Transport: STDIO, HTTP
Auth: Required
Registry: MCP registry, active
Tools verified: Nov 6, 2026

Related AI & LLM Tools MCP servers

View all

agentmemthezenmonster/agentmemGoverned memory for coding agents. Trust lifecycle, conflict detection, health scoring.2stars

Mcp Thinkgatetjp2021/mcp-thinkgateAuto-classifies prompt complexity and routes to the right Claude thinking mode. No API key needed.2stars

Pseudonym Mcpwoladi/pseudonym-mcpLocally pseudonymizes sensitive text before cloud LLM work and restores the tokens afterward.2stars

Pop Pay100xpercent/pop-payRuntime security for AI agent commerce. CLI + MCP server blocks hallucinated purchases.1stars

Forgevoxellinc/forge-mcpMCP server for Forge, Voxell's embedding API. Voxell's Ingot-8B-R3 ranks #1 for English on MTEB.1stars

Amberambermem/amberLong-term memory for AI assistants: permanent, portable across clients, searchable and exportable.1stars