CHANGELOG

What shipped.

Every release, in plain language. Desktop builds auto-update; the console and API deploy continuously.

v0.3.1
2026-08-29

The Solo plan: RepCue for one

Until now the smallest RepCue workspace was three seats. If you're a founder doing your own sales, that math never worked. It does now.

Solo: $29 a month, one seat

The full live copilot for exactly one person: sub-second suggestion cards grounded in your own documents, live transcript, post-call summary and follow-up email, CRM write-back, roleplay practice, your own analytics. 12 live hours a month included (about three calls a week), $2.50 an hour after that.

A spend cap instead of a surprise

Solo workspaces start with a $50 monthly overage cap. At the cap, new calls pause until renewal (a call already in progress always finishes), and you can raise, lower or remove the cap in Settings → Billing. Personal cards should never be surprised.

Built to grow out of

Solo is one seat by design; teams, coaching workflows, scorecard templates and scheduled reports stay on Team. When you hire, upgrade in place; your calls, knowledge base and deal history come with you.

v0.3.0
2026-08-21

Teams, role-based access, goals and deal risk

RepCue used to show everyone the same console. This release makes the product understand who you are and what you're accountable for.

Teams and the org chart

Create teams, assign a manager, add members, or push them from your identity provider. Any SCIM group that isn't admins, managers or reps now becomes a team with its memberships, so an IdP that already models your sales organisation configures RepCue's access control with no second setup.

Role-based read access

Every list (calls, deals, analytics, coaching, usage, search) is now scoped by role, and the server decides the scope:

  • A rep sees their own calls and deals. Always, whatever the client asks for.
  • A manager sees the teams they manage.
  • An admin sees the organisation.

Fetching a single record that's out of scope returns 404 rather than 403, so a response can't confirm that a colleague's call exists.

Managers in organisations that haven't created teams yet keep org-wide access, so nothing goes dark on upgrade. Creating your first team is what switches it on.

Goals with real pacing

Set a target for a person, a team or the org, over a week, month or quarter. RepCue judges it against what should have been achieved by now, not against the raw total, so 40 of 100 calls is on track on day 15 and behind on day 27. Rate metrics like adoption expect the full target from day one, and talk ratio is scored inverted, because lower is better.

A goal with no underlying data reports unknown. It is never reported as behind: a rep who hasn't taken a call hasn't missed their number.

Deal risk and forecast

Open deals now carry a 0–100 risk score built from what actually happened on the calls: staleness, missing or overdue next steps, slipped close dates, thin discovery, single-threading, competitor mentions, recurring objections. Every point is traceable to a named reason.

Signals with no underlying data are omitted, not scored as zero: a deal with no calls isn't safe, it's unmeasured, and each score reports its confidence so you can tell the difference. There's also a weighted forecast by stage and close month.

Coaching that keeps its receipts

Managers can now review and override any scorecard dimension with a rationale. The AI's original scorecard is never modified (reading a call returns both), so coaching disagreement stays auditable and the model's calibration stays measurable.

Admins can define custom scorecard rubrics of up to twelve dimensions, assigned per team or as the org default.

Notifications and digests

In-app, email and Slack notifications across four categories (coaching, pipeline, system and billing) with per-person preferences and a daily or weekly digest. Muting a category silences it everywhere, recurring alerts fire once per period (enforced in the database, not in memory), and an empty digest is never sent.

Per-team policies, with a ratchet

Consent, recording, screen-share exclusion, language and live coach-alert thresholds can now differ per team. One rule is absolute: a team may make consent stricter but never looser. If the org turns recording off, no team can turn it back on.

Also in this release

  • Custom vocabulary: product names, competitors and internal shorthand, applied to transcription and prompts.
  • Global search across calls, deals, people and documents, scoped like everything else.
  • Saved views (personal or shared) and scheduled reports delivered as CSV or a readable summary.
  • Admin health summary: seats, pooled-hour burn, integration health, stale knowledge, webhook failures, latency and consent coverage. Any block that can't be computed reports nothing rather than a misleading zero.
  • Read-only "view as": an admin can see the product as another user for 15 minutes. Mutations are refused at the authentication layer and every request is audited under both identities.

New documentation: Roles & permissions, Teams & org chart, Goals & targets, Coaching & scorecards, Deal risk & forecast, Notifications & digests, Per-team policies, Custom vocabulary, Saved views & reports.

v0.2.0
2026-08-21

Integrations, Windows, recordings, live coaching

The biggest release since the founding cohort opened. Everything below is in the 0.2.0 desktop build and the console/API that went live the same day.

Integrations, for real

  • CRM connected over OAuth. Salesforce, HubSpot and Pipedrive connect from Settings → Integrations with a normal consent screen, no pasted tokens. Deal context flows into the call pack; after the call, the proposed field changes ship to the CRM in one click from the review sheet or the desktop post-call view. Nothing is written without a human approving it.
  • Follow-up email to the prospect. Connect Google or Microsoft and the drafted follow-up sends as the rep through Gmail or Outlook. Without a mailbox connection it sends from RepCue with reply-to set to the rep.
  • Calendar auto-arm + pre-call brief. Google Calendar and Microsoft 365 calendars arm the HUD two minutes before each meeting and show a one-screen brief: matched deal, last call summary, open action items, the objections this account has raised before, and the battlecards most likely to matter.
  • Slack posts post-call summaries to deal channels. Recall meeting bots can now be dispatched to a meeting link for machines where the desktop app can't be installed.
  • Knowledge connectors. Import from Notion, Google Drive and Confluence, or index any URL. Stale documents finally have a fix button.

Desktop

  • Windows build. Signed NSIS installer for Windows 10/11 (x64), with silent install for MDM. macOS now ships Apple Silicon and Intel.
  • Live transcript panel with both sides, scrollback and in-call search.
  • Live talk-ratio + monologue coaching. A quiet nudge when you've been talking for 90 seconds straight, when your share climbs past 70%, or when a question lands in silence.
  • Discovery checklist. Budget, authority, need, timeline, competition and next step fill in live as the customer says them, and you're nudged when one is still missing late in the call.
  • Multilingual transcription. Pick the call language per call or per org; auto-detect is available.
  • Rep notes and snooze. Jot a note without leaving the HUD (it lands in the post-call sheet). Snooze suggestions for a few minutes when you just want the transcript.
  • Card feedback. Thumbs up/down with a one-tap reason. Unhelpful cards are down-ranked for the rest of the call and feed per-org tuning.
  • Post-call sheet edits and one-click ship from the desktop: approve CRM diffs, edit the email, send, done.
  • Roleplay mode. Practise against an AI customer that argues from your pricing and playbook, with a score at the end. Also used by the battlecard editor's Test in roleplay.
  • Remappable hotkeys with conflict warnings, update progress with a restart prompt, and opt-in crash reporting.

Console

  • Search across calls: full-text over transcripts, summaries and emails with highlighted snippets.
  • Recording and playback where the consent policy allows it; gametapes now play. Assist-only calls still store no audio, ever.
  • Integrations, Audit log, Webhooks, API keys, SSO/SCIM settings pages.
  • People management: change roles, deactivate, revoke and resend invites.
  • Playbooks and Deals pages, including a per-deal timeline of calls, artifacts and CRM changes.
  • Analytics: per-rep leaderboard, date-range filters, CSV export on every chart.
  • Coaching: comments on transcript turns, shareable snippet links, scorecard detail with rationale.
  • Invoice history, org data export, and right-to-erasure for users and organisations.

Platform

  • Public API with rk_live_… keys and OpenAPI docs; outbound webhooks signed with X-RepCue-Signature.
  • Cross-provider LLM failover so one provider's outage doesn't take suggestions down.
  • Real semantic embeddings on the live retrieval path.

Known limits in this release: recordings require a retention policy that allows storage; Pipedrive is token-based rather than OAuth; the Rust native audio taps remain in development (capture uses the documented OS loopback APIs today).

v0.1.0
2026-08-16

Founding cohort release

The first build handed to pilot teams.

  • Desktop HUD for macOS (Apple Silicon): sub-second suggestion cards grounded in your documents, with source and freshness on every card.
  • Dual-channel capture (your mic + system audio) so speaker separation is structural, plus a speakerphone mode for phone calls through the laptop mic.
  • HUD hidden from screen share by default, with an org-level policy to disable it.
  • Seven-step onboarding: sign-in, permissions, consent acknowledgement, roleplay preview, knowledge, integrations preview, ready.
  • Post-call summary, action items, MEDDIC/BANT extraction, drafted follow-up email, coaching scorecard.
  • Console: command center, live floor with manager listen-in and private whisper, call library with inline suggestions, knowledge base upload (PDF/DOCX/MD/TXT), battlecard editor with versioning, analytics, coaching, settings.
  • Jurisdiction-aware consent engine with assist-only mode; per-org retention windows.
  • Card-required self-serve signup, pooled live hours, metered overage.