deep_research_heavy
Long-form DR via gpt-5-5-pro + connector (5–30 min, monthly quota). Configurable via [models].heavydr
How to use it
deep_research_heavy is exposed by the Gpt2agent MCP server. Add the server to your MCP client (Claude Desktop, Cursor, Windsurf and others), and the deep_research_heavy tool becomes available to the model automatically. See the full listing for setup details and every tool this server provides.
Install Gpt2agent
pipx install gpt2agentOther tools in Gpt2agent (24)
Plan, country, groups, feature count, subscription expiry
Agent Mode — 262K context with autonomous browsing, code execution, tool use
Execute code via ChatGPT's Canvas feature (live editing environment)
Talk to any model on your account (gpt-5-3 default, override via model=). Pass gpt-5-5-pro, o3-pro, gpt-5-4-thinking, …
Run Python in ChatGPT's sandbox. Returns output + any generated charts/images
Kick off a new Codex task (resolves env from repolabel)
Read your current aboutuser / aboutmodel
Update them (read-modify-write, preserves unspecified fields)
Web-augmented research with citations (~30–120 s). Auto-confirms by default
Generate images via ChatGPT's built-in DALL-E. Returns download URLs + metadata
Full message history for a specific conversation (multimodal, code, images)
Temporary download URL for a ChatGPT file (~1h expiry)
Metadata for any ChatGPT file (images, uploads)
Talk through one of your private Custom GPTs (g-p-) — experimental
Connected apps + connectors
Codex environments (label, repos, network policy)
Recent Codex tasks + status
Recent ChatGPT conversations (titles: emails/phones redacted)
Your private g-p- GPTs
All models on your account (slug, maxtokens, reasoningtype, capabilities, enabledtools)
Scheduled / completed ChatGPT tasks
Add a memory (model-initiated workaround — POST /memories is 405)
List all ChatGPT memory entries (emails/phones redacted)
Keyword filter over memories