---
title: "AgentDrive — start in any agent, finish in any other"
description: "The shared workspace that lets Claude Code, Codex, OpenCode, Cursor and Antigravity pick up each other's sessions, with provenance on every artifact."
canonical: "https://agentdrive.sh/"
---
# AgentDrive

> AgentDrive is a shared workspace for coding agents that syncs each session from Claude Code, Codex, OpenCode, Cursor and Antigravity in the background, so any of those agents can resume work another one started, with provenance on every artifact and storage you can move to your own bucket.

Hooks in each of those five agents sync sessions in the background: a detached `adrive sync` uploads the redacted transcript, plus files the agent wrote that git does not track, to `sessions/<project>/<harness>/<session>.md` in the workspace. Any of them can resume another's session through the `drive_sessions` and `drive_resume` MCP tools, `/agentdrive:resume` in Claude Code, or `adrive resume` in a shell. The MCP server exposes 22 tools (18 stable plus 4 experimental sandbox tools) for versioned, content-addressed files, semantic search, run logs and sharing. Cline, Zed and Continue use the workspace tools as plain MCP clients, without sync. Bring your own bucket (R2 or S3) on the Team plan. Private Workspace encrypts the transcripts and files the CLI syncs before upload (MCP writes are not encrypted); redaction is best-effort and not a security boundary.

Install: `curl -fsSL agentdrive.sh/install | sh`

The page links below open Markdown directly. Every docs page is served as markdown at its URL plus `.md`, or at its normal URL with `Accept: text/markdown`. Start with the facts page for canonical answers, or read llms-full.txt for every docs page in one file.

## Facts

- [AgentDrive facts](https://agentdrive.sh/facts/index.md): Canonical AgentDrive facts: what it is, who makes it, supported agents, resume, local mode, pricing, license, self-hosting, privacy and official links.
- [Full documentation in one file](https://agentdrive.sh/llms-full.txt): The facts page and every docs page as markdown

## Docs

- [AgentDrive CLI (adrive)](https://agentdrive.sh/docs/cli.md): Install the adrive command-line tool, sign in, and run every command with its flags.
- [Content-addressing](https://agentdrive.sh/docs/concepts/content-addressing.md): How AgentDrive stores artifacts by the sha256 of their content, so identical content shares storage and history is immutable, and what that means for agents.
- [Cross-agent identity](https://agentdrive.sh/docs/concepts/cross-agent-identity.md): One API key is one principal across many agent platforms. Platform and role headers on each request give every session its own provenance in AgentDrive.
- [Memory and search](https://agentdrive.sh/docs/concepts/memory-and-search.md): How AgentDrive embeds artifacts on write, how drive_search, drive_recall and drive_context retrieve them by meaning, and why they are separate tools.
- [Sessions](https://agentdrive.sh/docs/concepts/sessions.md): How AgentDrive syncs coding-agent sessions in the background and lets another agent resume them, with privacy controls, Private Workspace and per-agent notes.
- [Workspaces](https://agentdrive.sh/docs/concepts/workspaces.md): The workspace is the unit of isolation in AgentDrive: when to use one or many, what lives inside one, and why a file is identified by its hash, not its path.
- [Data residency](https://agentdrive.sh/docs/data-residency.md): Where each AgentDrive data class lives, what BYO-bucket moves to your own R2 or S3 bucket in v1, what stays on AgentDrive, and what we commit to for v2.
- [AgentDrive Docs](https://agentdrive.sh/docs.md): Guides and reference for AgentDrive, the shared workspace where Claude Code, Codex, OpenCode, Cursor and Antigravity pick up each other's sessions.
- [License](https://agentdrive.sh/docs/license.md): How AgentDrive is licensed: Apache-2.0 for the MCP server, CLI, SDK and Claude Code plugin, AGPL-3.0 for the server, and a commercial license on request.
- [AgentDrive MCP server and tools](https://agentdrive.sh/docs/mcp-tools.md): Every tool the AgentDrive MCP server exposes today, grouped by sessions, files, collaboration, provenance, memory and runs, and which tool to choose when.
- [Quick start](https://agentdrive.sh/docs/quick-start.md): Install AgentDrive, initialize a project and make your first handoff between coding agents, with local mode, manual MCP config and raw REST as fallbacks.
- [AgentDrive REST API](https://agentdrive.sh/docs/rest-api.md): The AgentDrive REST API that MCP, the CLI and the SDKs wrap: authentication modes, identity headers, endpoints, search and runs, errors, rate limits, idempotency and versioning.
- [AgentDrive TypeScript SDK](https://agentdrive.sh/docs/sdk.md): The @adrive/sdk client for calling AgentDrive from TypeScript and JavaScript without MCP.
- [Self-hosting (community-supported)](https://agentdrive.sh/docs/self-hosting.md): Run your own AgentDrive API on your own Cloudflare account with one command: prerequisites, what the deploy script does, what is not included, and upgrades.
- [Roadmap](https://agentdrive.sh/docs/v1-roadmap.md): What's already shipped in AgentDrive, what's in flight next, what does not change, and what is parked: a stable summary updated when the scope changes.
- [Docs index](https://agentdrive.sh/docs/llms.txt): This docs section on its own

## Agent discovery

- [Agent protocol endpoints](https://agentdrive.sh/docs/mcp-tools.md): Public docs MCP, A2A, installable skills, catalogs and Web Bot Auth keys

## Product

- [Home](https://agentdrive.sh/index.md): This index plus the homepage questions, served at the homepage URL for markdown clients
- [Pricing](https://agentdrive.sh/pricing.md): Free $0 forever; Pro $20 per month; Team $30 per seat / month, 3-seat minimum; Enterprise Custom annual contract
- [Integrations](https://agentdrive.sh/integrations.md): Which agents sync sessions and which use the workspace tools
- [Claude Code integration](https://agentdrive.sh/integrations/claude-code.md): Plugin from the agentdrive-sh/agentdrive-mcp marketplace. Syncs sessions; resume with /agentdrive:resume.
- [Codex integration](https://agentdrive.sh/integrations/codex.md): Hooks, MCP server and prompts in ~/.codex. Setup approves the hooks for you.
- [OpenCode integration](https://agentdrive.sh/integrations/opencode.md): A plugin in ~/.config/opencode/plugins plus the MCP server. Resume with /agentdrive-resume.
- [Cursor integration](https://agentdrive.sh/integrations/cursor.md): Hooks in ~/.cursor/hooks.json and the MCP server in ~/.cursor/mcp.json. Resume with /agentdrive-resume.
- [Antigravity integration](https://agentdrive.sh/integrations/antigravity.md): Hooks, the MCP server and skills in ~/.gemini/config, shared by the agy CLI, the app and the IDE. Ask it to resume.
- [Security](https://agentdrive.sh/security.md): What AgentDrive syncs, what Private Workspace encrypts, and what stays managed when you bring your own bucket
- [Service status](https://agentdrive.sh/status.md)
- [About](https://agentdrive.sh/about.md)
- [Contact](https://agentdrive.sh/contact.md): Support, security, privacy and legal contacts

## Resume guides

- [Resume a session in another coding agent](https://agentdrive.sh/resume.md)
- [Resume a Claude Code session in Codex](https://agentdrive.sh/resume/claude-code-to-codex.md)
- [Resume a Claude Code session in OpenCode](https://agentdrive.sh/resume/claude-code-to-opencode.md)
- [Resume a Claude Code session in Cursor](https://agentdrive.sh/resume/claude-code-to-cursor.md)
- [Resume a Claude Code session in Antigravity](https://agentdrive.sh/resume/claude-code-to-antigravity.md)
- [Resume a Codex session in Claude Code](https://agentdrive.sh/resume/codex-to-claude-code.md)
- [Resume a Codex session in OpenCode](https://agentdrive.sh/resume/codex-to-opencode.md)
- [Resume a Codex session in Cursor](https://agentdrive.sh/resume/codex-to-cursor.md)
- [Resume a Codex session in Antigravity](https://agentdrive.sh/resume/codex-to-antigravity.md)
- [Resume an OpenCode session in Claude Code](https://agentdrive.sh/resume/opencode-to-claude-code.md)
- [Resume an OpenCode session in Codex](https://agentdrive.sh/resume/opencode-to-codex.md)
- [Resume an OpenCode session in Cursor](https://agentdrive.sh/resume/opencode-to-cursor.md)
- [Resume an OpenCode session in Antigravity](https://agentdrive.sh/resume/opencode-to-antigravity.md)
- [Resume a Cursor session in Claude Code](https://agentdrive.sh/resume/cursor-to-claude-code.md)
- [Resume a Cursor session in Codex](https://agentdrive.sh/resume/cursor-to-codex.md)
- [Resume a Cursor session in OpenCode](https://agentdrive.sh/resume/cursor-to-opencode.md)
- [Resume a Cursor session in Antigravity](https://agentdrive.sh/resume/cursor-to-antigravity.md)
- [Resume an Antigravity session in Claude Code](https://agentdrive.sh/resume/antigravity-to-claude-code.md)
- [Resume an Antigravity session in Codex](https://agentdrive.sh/resume/antigravity-to-codex.md)
- [Resume an Antigravity session in OpenCode](https://agentdrive.sh/resume/antigravity-to-opencode.md)
- [Resume an Antigravity session in Cursor](https://agentdrive.sh/resume/antigravity-to-cursor.md)

## Compare

- [Compare](https://agentdrive.sh/compare.md): How AgentDrive differs from the alternatives
- [AgentDrive vs Built-in resume](https://agentdrive.sh/compare/built-in-resume.md): Claude Code and Codex each resume their own sessions. AgentDrive syncs sessions so a different agent, or the same agent on another machine, can pick up the work.
- [AgentDrive vs Session history tools](https://agentdrive.sh/compare/session-history-tools.md): SpecStory and Entire save coding-agent sessions as Markdown or Git history you can search, share and resume. AgentDrive syncs sessions to a shared workspace and briefs the next agent.
- [AgentDrive vs Mem0](https://agentdrive.sh/compare/mem0.md): Mem0 is memory for AI agents and apps. AgentDrive is the shared workspace that lets coding agents resume each other's sessions.
- [AgentDrive vs Cloudflare Agent Memory](https://agentdrive.sh/compare/cloudflare-agent-memory.md): Cloudflare Agent Memory is managed memory for agents you build on Cloudflare. AgentDrive syncs and hands off sessions between the coding agents you already use.
- [AgentDrive vs Memory MCP](https://agentdrive.sh/compare/memory-mcp.md): The Memory MCP server keeps a local knowledge graph of facts an agent chooses to save. AgentDrive syncs whole coding-agent sessions so another agent can resume the work.
- [AgentDrive vs Filesystem MCP](https://agentdrive.sh/compare/filesystem-mcp.md): Filesystem MCP gives one agent access to folders on one machine. AgentDrive is a shared, versioned workspace every agent can reach.
- [AgentDrive vs CLAUDE.md and handoff notes](https://agentdrive.sh/compare/claude-md-handoff-notes.md): CLAUDE.md, AGENTS.md and handoff notes hold what you write down for the next session. AgentDrive keeps what each session did, so another agent can pick it up.

## Glossary

- [Glossary](https://agentdrive.sh/glossary.md): Short definitions of the terms behind coding agents and cross-agent work: MCP, AGENTS.md, CLAUDE.md, session sync, handoff briefs, provenance and more.
- [AGENTS.md](https://agentdrive.sh/glossary/agents-md/index.md): AGENTS.md is an open format for guiding coding agents: a markdown file, usually at a repository's root, that holds the instructions an agent needs to work in that project.
- [CLAUDE.md](https://agentdrive.sh/glossary/claude-md/index.md): CLAUDE.md is a markdown file of instructions that Claude Code reads at the start of every session, giving it persistent context for a project, your personal workflow or an organization.
- [Coding agent](https://agentdrive.sh/glossary/coding-agent/index.md): A coding agent is an AI tool that works through a software task in a code repository: it reads and edits files, runs commands such as tests, and reports back.
- [Content-addressed storage](https://agentdrive.sh/glossary/content-addressed-storage/index.md): Content-addressed storage keeps each piece of data under a key derived from its content, usually a cryptographic hash, so identical content is stored once and a stored object never changes.
- [Cross-agent resume](https://agentdrive.sh/glossary/cross-agent-resume/index.md): Cross-agent resume is continuing a coding task in a different coding agent from the one that started it, by loading the earlier session instead of explaining the task again.
- [Handoff brief](https://agentdrive.sh/glossary/handoff-brief/index.md): A handoff brief is the summary of a coding-agent session that AgentDrive gives the next agent: the goal, the files touched and the most recent turns, sized to a character budget.
- [Local mode](https://agentdrive.sh/glossary/local-mode/index.md): Local mode is AgentDrive without an account: the adrive CLI reads each agent's transcripts on your machine and builds the handoff brief there, so nothing is uploaded.
- [MCP server](https://agentdrive.sh/glossary/mcp-server/index.md): An MCP server is a program that provides context to AI applications over the Model Context Protocol, exposing tools, resources and prompts that a connected MCP client can discover and use.
- [Model Context Protocol (MCP)](https://agentdrive.sh/glossary/mcp/index.md): The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems such as data sources, tools and workflows.
- [Principal (AgentDrive)](https://agentdrive.sh/glossary/principal/index.md): In AgentDrive, a principal is the identity behind an API key: a human or a team that owns workspaces, holds the plan and is billed, whichever agents act on its behalf.
- [Private Workspace](https://agentdrive.sh/glossary/private-workspace/index.md): Private Workspace is an AgentDrive setting that makes the adrive CLI encrypt the session transcripts and files it syncs on your machine with AES-256-GCM before upload. MCP writes are not encrypted.
- [Provenance (agent provenance)](https://agentdrive.sh/glossary/provenance/index.md): Agent provenance is the record of which agent, on which platform and under which identity, created or changed an artifact, and when, so work stays attributable across agents and sessions.
- [Session sync](https://agentdrive.sh/glossary/session-sync/index.md): Session sync copies a coding-agent session from the machine it runs on into a shared workspace as it happens, so another agent or another machine can read and resume it.
- [Transcript adapter](https://agentdrive.sh/glossary/transcript-adapter/index.md): A transcript adapter is code that reads one coding agent's session transcripts from disk and turns each session into a common document of turns, working directory, title and files touched.
- [Usage limit (coding agents)](https://agentdrive.sh/glossary/usage-limit/index.md): A usage limit is the cap a coding agent's plan puts on how much you can use it in a time window; once you reach it, the agent pauses until the limit resets.

## Blog

- [Blog](https://agentdrive.sh/blog.md)
- [Blog index](https://agentdrive.sh/blog/llms.txt): This blog section on its own
- [AgentDrive architecture: the pieces and how they fit](https://agentdrive.sh/blog/agentdrive-architecture/index.md): The CLI and hooks, the stdio MCP server, one Hono Worker on KV, R2, Vectorize and Workers AI, the hosted MCP with OAuth, and the web app, end to end.
- [AGENTS.md, CLAUDE.md and session history: what each is for](https://agentdrive.sh/blog/agents-md-claude-md-session-history/index.md): AGENTS.md holds repo rules for every agent, CLAUDE.md and auto memory hold Claude Code's own, and session history holds the task. What goes in each layer.
- [Antigravity limits: continue your task in another agent](https://agentdrive.sh/blog/antigravity-usage-limit/index.md): Hit your Antigravity quota mid-task? What Google says about the limits, and how to move the same task into Claude Code, Codex or Cursor with AgentDrive.
- [Behind the scenes: how AgentDrive is built](https://agentdrive.sh/blog/behind-the-scenes-building-agentdrive/index.md): The stack and the decisions behind AgentDrive: KV and R2 with no Postgres, MCP first, an AGPL server, and the one founder building it with parallel agents.
- [What to do when Claude Code hits its usage limit](https://agentdrive.sh/blog/claude-code-usage-limit/index.md): Claude Code hit its usage limit mid-task? You can wait, pay for more usage or switch agents. Here is how to carry the same task into Codex or Cursor.
- [Codex to Claude Code, end to end](https://agentdrive.sh/blog/codex-to-claude-code-walkthrough/index.md): A step-by-step walkthrough of moving a Codex session into Claude Code with AgentDrive, from setup and sync to the handoff brief and checking files.
- [What to do when Codex hits its usage limit](https://agentdrive.sh/blog/codex-usage-limit/index.md): Codex hit its usage limit mid-task? You can wait for the reset, buy credits or switch agents. Here is how to carry the task into Claude Code or Cursor.
- [Content-addressed storage on R2 and KV](https://agentdrive.sh/blog/content-addressed-storage-r2-kv/index.md): How AgentDrive keeps file bytes at content/<sha256> in R2 and all metadata in KV, and what that costs in consistency, enumeration and missing SQL.
- [The real cost of switching coding agents](https://agentdrive.sh/blog/cost-of-switching-coding-agents/index.md): Switching coding agents costs you the re-explaining. The goal, the files touched, the decisions and the dead ends. What a handoff should carry and leave out.
- [Cursor and Claude Code without re-explaining](https://agentdrive.sh/blog/cursor-and-claude-code/index.md): Hand work between Cursor and Claude Code in both directions with AgentDrive, including the Cursor hook, the project-directory catch and /agentdrive-resume.
- [Out of Cursor requests? Keep going in another agent](https://agentdrive.sh/blog/cursor-usage-limit/index.md): Ran out of Cursor usage mid-task? Wait for the weekly reset, top up, or move the same task into Claude Code or Codex with a handoff brief from AgentDrive.
- [Reading five coding agents' transcript formats](https://agentdrive.sh/blog/five-agent-transcript-formats/index.md): How AgentDrive parses Claude Code, Codex, OpenCode, Cursor and Antigravity sessions into one TranscriptDoc, and how hooks fill in what each format leaves out.
- [Getting started with AgentDrive](https://agentdrive.sh/blog/getting-started-with-agentdrive/index.md): Install the adrive CLI, run adrive init signed in or in local mode, then make your first sync and your first resume between two coding agents.
- [The handoff brief format](https://agentdrive.sh/blog/handoff-brief-format/index.md): The AgentDrive handoff brief as a format: the session document it reads, its fields, order and budget, the MCP tools that serve it and the identity headers.
- [Handoff notes, CLAUDE.md or AgentDrive: which to use](https://agentdrive.sh/blog/handoff-notes-vs-claude-md/index.md): CLAUDE.md holds rules, a handoff note holds your judgment about one task, and AgentDrive holds what a session did. When each fits and how they combine.
- [Hooks that never block the agent](https://agentdrive.sh/blog/hooks-that-never-block/index.md): How AgentDrive's sync hooks fit inside the 3 seconds Codex allows a SessionEnd hook, with a detached sync, a lock file, a two-minute debounce and failures that retry quietly.
- [How AgentDrive works](https://agentdrive.sh/blog/how-agentdrive-works/index.md): The plain-language overview: sessions sync in the background, land in a workspace, and the next agent asks to resume and gets a short handoff brief.
- [Resume across agents with no account](https://agentdrive.sh/blog/local-mode-no-account/index.md): AgentDrive's local mode in depth. What adrive init --local sets up, where each agent's sessions are read from, which agents need hooks, and the limits.
- [One MCP server, with and without an account](https://agentdrive.sh/blog/mcp-server-local-and-hosted/index.md): How @adrive/mcp-server answers drive_sessions and drive_resume from local transcripts with no API key, and serves the same tool registry over hosted OAuth MCP.
- [Best practices for coding with more than one agent](https://agentdrive.sh/blog/multi-agent-coding-best-practices/index.md): Practical habits for moving a task between coding agents: a clear first message, commit before switching, check files after resuming, and when to stay put.
- [One task, three agents: OpenCode, Codex and Claude Code](https://agentdrive.sh/blog/opencode-codex-claude-code/index.md): How to pass one task between OpenCode, Codex and Claude Code with AgentDrive, how the OpenCode plugin syncs sessions, and how to pick the right session.
- [Privacy and safety guardrails in AgentDrive](https://agentdrive.sh/blog/privacy-and-safety-guardrails/index.md): Every AgentDrive privacy control, from local mode and the deny list to redaction, encryption, roles, share links and rate limits, and what each one misses.
- [How Private Workspace encrypts your sessions](https://agentdrive.sh/blog/private-workspace-encryption/index.md): How AgentDrive's Private Workspace encrypts synced sessions with AES-256-GCM on your machine, what the server still sees, and how to back up your keys.
- [Keeping secrets out of synced agent transcripts](https://agentdrive.sh/blog/secrets-in-agent-transcripts/index.md): What AgentDrive syncs from your agent sessions, which files it never uploads, exactly which secrets the redactor catches, and the habits that cover the rest.
- [Self-host AgentDrive on Cloudflare](https://agentdrive.sh/blog/self-host-agentdrive-on-cloudflare/index.md): What pnpm selfhost creates on your Cloudflare account, what it needs, how to connect the CLI with an API key, and what the AGPL asks of you in return.
- [Running two coding agents on one task without losing context](https://agentdrive.sh/blog/two-agents-one-task/index.md): When splitting one coding task between two agents pays off, and how a handoff brief keeps Claude Code, Codex or Cursor in step, with the exact commands.
- [How to use AgentDrive in Antigravity](https://agentdrive.sh/blog/use-agentdrive-with-antigravity/index.md): What adrive init writes under ~/.gemini/config for the agy CLI, app and IDE, the 30-minute idle rule, and how to resume another agent's session in Antigravity.
- [How to use AgentDrive in Claude Code](https://agentdrive.sh/blog/use-agentdrive-with-claude-code/index.md): Install the AgentDrive plugin in Claude Code, see what it syncs, resume a Codex or Cursor session with /agentdrive:resume, and keep going when you hit a limit.
- [How to use AgentDrive in Codex](https://agentdrive.sh/blog/use-agentdrive-with-codex/index.md): What adrive init writes into ~/.codex, how it approves the hooks, why they use a 3-second timeout, and how to resume a Claude Code session in Codex.
- [How to use AgentDrive in Cursor](https://agentdrive.sh/blog/use-agentdrive-with-cursor/index.md): What adrive init merges into ~/.cursor, why Cursor needs a hook to know your project, and how to resume a Claude Code or Codex session with /agentdrive-resume.
- [How to use AgentDrive in OpenCode](https://agentdrive.sh/blog/use-agentdrive-with-opencode/index.md): How the AgentDrive plugin syncs OpenCode sessions, how to resume a Claude Code or Codex session with /agentdrive-resume, and what to do when a provider stops.
- [What goes into a handoff brief, and what doesn't](https://agentdrive.sh/blog/what-goes-into-a-handoff-brief/index.md): An AgentDrive handoff brief carries a session's goal, files touched and recent turns in a fixed budget. Here is what it holds, what it leaves out and why.
- [Why AgentDrive's server is AGPL and its clients Apache-2.0](https://agentdrive.sh/blog/why-agpl-server-apache-clients/index.md): Why the AgentDrive server is AGPL-3.0 and its clients Apache-2.0, and what that means for users, self-hosters and contributors, the CLA and trademarks.
- [Why one model can't continue another's hidden state](https://agentdrive.sh/blog/why-model-state-doesnt-transfer/index.md): A coding agent's reasoning and prompt cache stay with its provider. What moves between agents is text and files, so a handoff should be a short, honest brief.
- [Write a transcript adapter for a new agent](https://agentdrive.sh/blog/write-a-transcript-adapter/index.md): A contributor guide to AgentDrive transcript adapters, from the TranscriptSource interface and the Claude Code reference to registration, tests and hooks.
- [How background session sync works](https://agentdrive.sh/blog/how-background-session-sync-works/index.md): How AgentDrive syncs agent sessions in the background: hooks in five agents, a detached sync, redaction and ignore rules, storage and the handoff brief.
- [Resume your Conductor sessions in any agent](https://agentdrive.sh/blog/moving-off-conductor-cloud/index.md): How to sync the Claude Code and Codex sessions you run in Conductor to AgentDrive, so any coding agent can resume them, alongside Conductor.
- [Start in Claude Code, pick up in Codex](https://agentdrive.sh/blog/start-in-claude-code-pick-up-in-codex/index.md): How AgentDrive syncs your coding-agent sessions in the background so Claude Code, Codex, OpenCode, Cursor and Antigravity can resume each other's work.

## Optional

- [Source code](https://github.com/agentdrive-sh/agentdrive): the AGPL server, web app, CLI, SDK, MCP server and Claude Code plugin, with AGENTS.md for coding agents
- [Manifesto](https://agentdrive.sh/manifesto.md): Why AgentDrive was built: a shared artifact workspace with provenance for coding agents
- [Legal](https://agentdrive.sh/legal.md)
- [Terms](https://agentdrive.sh/legal/terms.md)
- [Privacy](https://agentdrive.sh/legal/privacy.md)
- [Data Processing Agreement](https://agentdrive.sh/legal/dpa.md)

## Questions

### How do I install AgentDrive?

Run `curl -fsSL agentdrive.sh/install | sh`, which installs the `adrive` CLI and runs `adrive init` to sign you in through the browser and wire session sync hooks, the MCP server and the resume commands into Codex, OpenCode, Cursor and Antigravity, plus the Claude Code plugin when the `claude` command is on your PATH.

The CLI needs Node.js 22.15 or later. Read the script first with `curl -fsSL agentdrive.sh/install | less`. To set up Claude Code on its own, run `/plugin marketplace add agentdrive-sh/agentdrive-mcp` and `/plugin install agentdrive@agentdrive` inside it, then `/agentdrive:login`. Restart your agents afterwards so they load the new hooks.

### Which agents does AgentDrive work with?

Session sync and resume work in Claude Code, Codex, OpenCode, Cursor and Antigravity, and any of them can pick up a session another one started; Cline, Zed and Continue connect as plain MCP clients, so they can use the workspace tools and resume sessions, but their own sessions are not synced.

### What does the next agent get when it resumes?

It gets a handoff brief with the source agent and session details, the goal from the session's first message, the files touched, and as many recent turns as fit in 12,000 characters by default, and it is told to check the files on disk before relying on the brief.

Open the next agent in the same project directory and run `/agentdrive:resume` in Claude Code or `adrive resume` in a terminal, or ask the agent to resume your last AgentDrive session.

### Do I need an account?

No: local mode resumes sessions between agents on one machine without an account, by reading each agent's own transcripts on disk, so nothing is uploaded and local resumes are unlimited; an account adds sync across machines, a shared workspace, semantic search, share links and the provenance timeline.

Run `adrive init --local` to wire the hooks without signing in, then `adrive resume` in your project directory.

### What gets uploaded from my machine?

With an account, AgentDrive uploads session transcripts and the files an agent wrote that git does not track, after redacting common secret shapes, while `.env` files, keys, SSH and cloud credentials and anything in `.adriveignore` stay local, and `adrive privacy` switches transcript or file sync off.

Redaction is best-effort, and ignore rules cover files, not text you paste into a conversation. With Private Workspace on (`adrive privacy --private-workspace on`), the `adrive` CLI encrypts the session transcripts and files it syncs with AES-256-GCM before upload, and that content is not searchable by meaning. Files saved through MCP tools such as `drive_write` are not encrypted, `adrive watch --no-encrypt` and `adrive push --plaintext` skip encryption, and run metadata, storage paths and version messages are stored unencrypted.

### Does syncing slow my agent down?

No: the agent's hook starts a detached `adrive sync` and exits right away, so it never blocks the agent, and a failing hook never surfaces in your session; turn-end events re-sync a session at most every two minutes, while session-end and compaction events always sync.

Hook-started syncs log to `~/.config/agentdrive/sync.log`, and `adrive sync --dry-run` shows what a sync would pick up without uploading anything.

### Is AgentDrive open source, and can I self-host it?

Yes: the server is AGPL-3.0 and the CLI, MCP server, SDK, installer and Claude Code plugin are Apache-2.0, and `pnpm selfhost` deploys the same server to your own Cloudflare account, which needs the Workers Paid plan and runs API-key only, with no billing and no plan limits.

Self-hosting is community-supported. The source is at https://github.com/agentdrive-sh/agentdrive and the guide is at https://agentdrive.sh/docs/self-hosting.
