Try Sortd for FREETry SortdBoth Sortd and Drag now let AI agents work a Gmail shared inbox. The difference is the model. Drag bundles its own agent (Drag Agent, early access) and a proprietary “Intelligence” knowledge layer, metered in credits. Sortd opens the workspace to any agent — Claude, ChatGPT, Gemini, or whichever you already use — through a full API and MCP server at a flat seat price. The idea is simple: you already have an agent, and you shouldn't be locked into one company's. Hosted options like Claude Managed Agents set up in minutes.
Full API + MCP on Business · drafts-for-review by default · graduated send permissions · no credits, ever
| Sortd | Drag | |
|---|---|---|
| MCP access | Business $22 ($28) — full API + MCP | Plus $18 ($22) — MCP only, 2,500 credits/mo |
| Full API + MCP | Business $22 ($28) | Pro $24 ($30) — 10,000 credits/mo |
| Agent operation pricing | Flat seat. No credits, ever. Fair-use policy | Credits: read 1 · write 2 · send 5 · AI 15 — “free during early access”, later price unpublished |
| Agent model | Connect any agent: Claude, ChatGPT/Codex, Gemini, Cursor, Claude Managed Agents | Drag Agent (theirs) — Early Access |
| Knowledge layer | Your agent's own (Claude Skills, memory tools) — portable | Proprietary “Intelligence” — Early Access |
| Approval model | Drafts-for-review default; graduated scopes (draft → constrained reply → approved send); audit trail | Confidence-threshold auto-send (~85%) with escalation |
| Workflow surface for agents | Custom fields, CRM pipelines, SLA state, assignment — in the tool surface | Board/email primitives; deeper workflow on Pro |
| Setup | One API key on your existing workspace | API key on Pro |
Drag's bet is that you adopt their agent and their knowledge layer — both early access, both on a meter. Sortd's bet is the opposite: the world's best agents already exist, the AI companies invest billions a year improving them, and your business knowledge belongs in your agent — Claude Skills, memory, MD knowledge bases — where it stays portable to any future stack. Sortd gives that agent what it doesn't have: your team's real mailboxes, workflow state, approvals and an audit trail. Connect any agent. Sortd is the workspace where it works with your team.
Every read, write, send and AI action in Drag consumes credits that are only free “during early access” — the post-launch price is unpublished. Sortd's answer is structural: your agent runs on your own model subscription, so there is nothing for us to meter. Flat seats, a plain-English fair-use policy, done.
On Sortd, agents draft by default. An admin grants scopes deliberately: constrained reply (recipients derived server-side, an agent can never invent an address), approve-by-sending, and full send only where you explicitly allow it. Every agent action is logged. Escalations land assigned to a named human on the board — and the agent resumes routine follow-up after the human steps in.
Claude Desktop, Claude Code, ChatGPT, Gemini, Cursor — any MCP client connects in minutes, and hosted Claude Managed Agents run with no infrastructure at all. Your agent also brings its other tools into the same session: Stripe, Notion, Slack, your CRM — we didn't build those integrations; your agent brings them. Comparing the team inboxes themselves? Sortd vs Drag for teams →
Is Drag's agent available today?
Drag Agent and the Intelligence layer are Early Access as of July 2026.
What do Drag credits cost after early access?
Unpublished as of July 2026. Sortd has no meter.
Which agents work with Sortd?
Any MCP client or REST caller — Claude, ChatGPT/Codex, Gemini, Cursor, Copilot, LangChain agents, hosted Claude Managed Agents.
Can the agent send email?
Only under scopes you grant. Drafts-for-review is the default; full send is an explicit opt-in.
Where teams and agents reply together · 7-day free trial · no credit meters
Pricing verified July 2026 on dragapp.com/pricing and sortd.com/pricing. See all comparisons →