GPT2AgentBLOCK
Your codex login → a full ChatGPT Plus/Pro account (every model, deep research, image gen, code exec) inside Claude Code, Codex & any MCP client. One-line install.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
MCP server for your ChatGPT account: `codex login` → ChatGPT Plus/Pro inside any MCP client.
An MCP server that puts your ChatGPT Plus or Pro subscription — Chat models and the account-tier features below — inside Claude Code, Codex, Cursor, Windsurf, Zed, and any MCP client.
[](https://pypi.org/project/gpt2agent/) [](https://pepy.tech/projects/gpt2agent) [](https://pypistats.org/packages/gpt2agent) [](https://github.com/robotlearning123/gpt2agent/actions/workflows/ci.yml) [](./LICENSE) [](https://pypi.org/project/gpt2agent/)
📖 Quickstart · Client setup · How it works · Troubleshooting · FAQ · Docs index · Account safety
TL;DR
pipx install gpt2agent # 1. install codex login # 2. authenticate (or: gpt2agent setup) gpt2agent install # 3. register with your MCP client(s), then restart it gpt2agent doctor # 4. verify — status table, never spends quota
Then ask your agent to call chat, deep_research, or account_status.
One thing to know up front — conversation tools have three lanes (below): REST needs the sentinel bridge (owner-supplied, not distributed); browser needs a one-time Chrome login; manual works with zero network. Read-only tools work out of the box.
What works right now
September 29 update:
43b27312d3ddOBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add gpt2agent -- None gpt2agent==0.0.26 run --stdio
Exposed tools (38)
28 read · 8 write · 2 destructive. Blast radius: 2 tools can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
account_status | read | Return ChatGPT account info. |
agent | read | ChatGPT Agent Mode — 262K context with autonomous browsing, code |
canvas_execute | write | Execute code through ChatGPT |
chat | read | Chat with a ChatGPT Chat model exposed by your account. |
code_interpreter | write | Execute code via ChatGPT |
codex_task_create | write | Create a new Codex task. |
create_automation | write | Create a scheduled automation (dot recurring work). Defaults to DISABLED. |
custom_instructions_get | read | Return ChatGPT custom instructions (PII redacted). |
custom_instructions_set | destructive | Overwrite ChatGPT custom instructions (read-modify-write — preserves fields not supplied). |
deep_research | read | Search the web and synthesize a detailed report with citations. |
deep_research_heavy | read | Long-form Deep Research using gpt-6-pro (5–30 min, uses monthly DR quota — check /backend-api/conversation/init for remaining). For short web-augmented answers use `deep_research` instead. |
dot_messages | read | Read the dot conversation (newest last). |
dots_status | read | Report whether OpenAI dots (always-on agents) are usable on this account. |
generate_image | read | Generate an image using ChatGPT |
get_conversation | read | Get full details of a ChatGPT conversation including all messages. |
get_file_download_url | read | Get a temporary download URL for a ChatGPT file. |
get_file_info | read | Get metadata for a ChatGPT file (images, uploads, etc.). |
gpt_chat | read | Chat through one of your private Custom GPTs. |
list_apps | read | Return ChatGPT connected apps/connectors. Names unresolvable — IDs with type classification returned. |
list_automations | read | List scheduled automations (the dot |
list_codex_envs | read | Return Codex environments (label, repos, network access). |
list_codex_tasks | read | Return recent Codex tasks (title + status). Content is PII-redacted. |
list_conversations | read | Return recent ChatGPT conversations (titles PII-redacted). |
list_custom_gpts | read | List your private Custom GPTs from the ChatGPT sidebar. |
list_dots | read | List the account |
list_models | read | List the model catalog exposed by your account. |
list_tasks | read | Return scheduled/completed ChatGPT tasks with full metadata (titles PII-redacted). |
memory_create_via_chat | write | Add an entry to your ChatGPT memories. |
memory_list | read | Return all ChatGPT memories (PII redacted). |
memory_search | read | Keyword search over ChatGPT memories. Returns matching entries (PII redacted). |
queue_cancel | read | Cancel a queued/waiting task. Already-terminal or running tasks |
queue_result | read | Fetch a finished task |
queue_status | read | Show one task (`task_id`) or the whole queue when omitted. |
queue_submit | write | Enqueue work on this host |
remove_automation | destructive | Delete a scheduled automation (id from `list_automations`). |
send_to_dot | write | Send a message to your dot (async delivery; the dot replies later). |
set_automation_status | write | Enable or disable a scheduled automation (id from `list_automations`). |
usage_stats | read | Account usage snapshot: per-model caps, per-feature remaining |
Trust audit
BLOCKgrade D · trust 65/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (5 observation(s))
- Network
- declared (3 observation(s))
- Shell
- declared (1 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (22)
err " Debian/Ubuntu: sudo apt install pipx && pipx ensurepath"
err " Fedora: sudo dnf install pipx && pipx ensurepath"
return importlib.import_module(f"wrapper.reverse.{name}")random.choice(navigator_key),
random.choice(document_key),
random.choice(window_key),
random.choice(_WINDOW_KEYS),
token = "eyJhbGciOi.eyJzdWIiOi.c2lnbmF0dXJl"
secret = "sk-ABCDEFGHIJKLMNOPQRST"
api_key = "sk-ABCDEFGHIJKLMNOPQRST"
custom_instructions_set, remove_automation
.migration-v1-complete
src = inspect.getsource(importlib.import_module(modname))
target = bytes.fromhex(diff)
decoded = base64.b64decode(dx)
pm[e] = base64.b64decode(_to_str(pm[e])).decode()
run: sudo apt-get update -qq && sudo apt-get install -y -qq libcurl4-openssl-dev
run: sudo apt-get update -qq && sudo apt-get install -y -qq libcurl4-openssl-dev
mcp, curl_cffi, uvicorn, tomli
curl -fsSL https://raw.githubusercontent.com/robotlearning123/gpt2agent/main/install.sh | bash
curl -fsSL https://raw.githubusercontent.com/robotlearning123/gpt2agent/main/install.sh | bash
system use found in code, not declared in the description
Gates applied: no_behavioural_pass.
43b27312d3ddfull audit observations/trust-audit/mcp-server/robotlearning123__gpt2agent.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | 43b27312d3dd | BLOCK | D | 65 | first audit |
Questions
What is the GPT2Agent MCP server?
Your codex login → a full ChatGPT Plus/Pro account (every model, deep research, image gen, code exec) inside Claude Code, Codex & any MCP client. One-line install.
What tools does GPT2Agent expose?
38 in total: 28 read-only, 8 that write, and 2 that can delete or overwrite (custom_instructions_set, remove_automation). Every one is listed on this page with its risk.
Is GPT2Agent safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (65/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 2 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does GPT2Agent need?
No credential environment variables were found in its source, so it appears to need none.
How does GPT2Agent run?
It speaks stdio and streamable-http, so it runs as a local process your client starts. It is published on PyPI as gpt2agent.
How current is this page?
The grade is for one exact copy of the source (43b27312d3dd), read on 2026-10-08. The repository is watched and re-audited when it changes.