august
by august
Interview simulated customer twins, run focus groups and field surveys from an AI agent. Anthropic lists 32 tools; august publishes no tool list. OAuth sign-in with identity-only scopes, and every research run spends metered credits.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect august via MCP
https://mcp.getaugust.ai/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
august Tools & Capabilities (32)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- The OAuth scopes are identity-only.
login.getaugust.aiadvertisedemail,offline_access,openidandprofileon 2026-08-22, and the server's RFC 9728 descriptor declares noscopes_supportedat all. Those four are OpenID identity claims, not permissions — they read like a scope list but draw no boundary. No boundary falls between reading a transcript and spending $50 on a survey. - No tool reads the credit balance. Every billing operation is exposed and no accounting one is. This is the single most consequential gap in the listing.
- august publishes no tool list. The 32 names come from Anthropic's directory. We could not diff them against a vendor enumeration because none exists on august's site, npm, PyPI or GitHub.
- We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail is published here, and we cannot say which tools accept free text versus structured input.
- Twins are predictions, not customers. august's own evaluation post reports a twin picking a held-out real session out of four candidates 66% of the time against 25.5% for default Claude — a real gap, and also a 34% error rate. august's designer writes that "real customer conversations are what confirm it". Do not treat a focus group result as a finding.
- Custom twins need your behavioral data. august states twins are trained from existing session recordings, event data, support platforms, CDPs or a warehouse, or from august's own pixel, and that training time varies with how active your users are. Out of the box you get generic vertical twins, not twins of your users.
- The directory tier and august's own docs disagree. Anthropic's directory records august at
partnertier; august's install guide, published 2026-07-08, calls itself "a community connector in Claude's directory". We report Anthropic's tier because the tier is Anthropic's to assign, but the vendor's own page has not caught up. - Panel sizes are capped by account. In august's GitHub App — a different product, but the same twin panel underneath — a panel above three is refused as above the account's limit unless raised. Whether
select_twinshits a comparable ceiling is not documented.
Frequently asked questions
Yes, every research run spends credits. august prices credits at five cents each and estimates about 20 credits for a one-to-one conversation, about 250 for a focus group, and 800 to 1,200 for a survey. So an agent that starts a survey spends roughly 40 to 60 dollars of a metered balance in one call, at august's own published rates.
No. None of the 32 tools Anthropic lists reads a credit balance, a usage figure or a quota. The tools that spend credits are all present, and the one that would tell you what spending them costs you is not. Check your balance in the august web app before letting an agent run focus groups or surveys unattended.
Only identity claims. august's authorization server advertised four scopes on 2026-08-22 — email, offline_access, openid and profile — and none of them names a research operation. The server's RFC 9728 descriptor declares no scopes_supported at all. Nothing at the consent screen separates reading a transcript from starting a survey.
Thirty-two, according to Anthropic's directory. august publishes no tool list of its own anywhere we could find, so there is no second enumeration to check that number against. Its install guide names capabilities in prose — interviews, focus groups, results — but never a tool name, so the count is unverified rather than confirmed clean.
Only one tool carries a removal verb, and it is reversible in name. archive_segment archives a saved group of twins rather than deleting it. No tool among the 32 names a delete of an interview, focus group, survey or organization. end_focus_group and cancel_survey stop a run in progress rather than erasing what it produced.
No. Every participant is a simulated twin, not the person it models. august states each twin is a digital representation built from anonymized behavioral session data, and the interviews, focus groups and surveys run entirely inside august. No tool among the 32 names an email, message, notification or other outbound channel to a real individual.
A paid Claude plan and an august account. august's install guide states connectors are available on Claude's Pro, Max, Team and Enterprise plans, and that the first connection asks you to sign in with an august account you create at app.getaugust.ai. Some organizations only let owners and admins add connectors at all.
Behavioral session history, not identity. august states no personally identifiable information is captured, only behavioral and site-navigation signals, kept as long as the account stays active. Each twin stores dated memories of what its real counterpart did on your product, plus reflections drawn from them, and an agent reading a transcript sees that recalled detail.
Sources
- august MCP install guide (retrieved 2026-08-22). The page renders 725 words as plain text; the setup detail for Claude, Claude Code, Cursor and ChatGPT is all in that text, and its Next.js RSC flight payload holds no additional content. · retrieved 2026-08-22
- august homepage, including the four-tab FAQ and the pricing and credit-usage tables (retrieved 2026-08-22). The FAQ answers and credit estimates are client-rendered and appear only in the page's JavaScript route bundle, not in the served HTML. · retrieved 2026-08-22
- august blog, "Introducing august" (retrieved 2026-08-22) · retrieved 2026-08-22
- august blog, "How we train Digital Twins" (retrieved 2026-08-22) · retrieved 2026-08-22
- august blog, "How we evaluate digital twins" (retrieved 2026-08-22) · retrieved 2026-08-22
- august blog, "How I use august as a designer" (retrieved 2026-08-22) · retrieved 2026-08-22
- august blog, "How to install our GitHub App" (retrieved 2026-08-22) · retrieved 2026-08-22
- august sitemap, eight URLs total (retrieved 2026-08-22).
robots.txtis a blanket allow; there is nollms.txt. · retrieved 2026-08-22 - Live auth check: anonymous initialize to
https://mcp.getaugust.ai/mcpreturned 401 with aWWW-Authenticate: Bearerchallenge naming its RFC 9728 descriptor, read athttps://mcp.getaugust.ai/.well-known/oauth-protected-resource/mcp; authorization server metadata athttps://login.getaugust.ai/.well-known/oauth-authorization-server(2026-08-22) · retrieved 2026-08-22 - Fermàt Commerce privacy policy (retrieved 2026-08-22) · retrieved 2026-08-22
- Anthropic Connectors Directory entry — , read from our directory snapshot dated 2026-08-16 · retrieved 2026-08-16
- august support — <mailto:support@mail.getaugust.ai>
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Analytics
- Developer
- august
- Tools
- 32
- Domain
- mcp.getaugust.ai
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.