PerfCopilot
Get started

Integrations overview

PerfCopilot grounds reviews in evidence pulled from the tools your team already uses. Each integration is a separate signal source — a stream of facts about what someone did in a cycle.

This page covers what we pull from each integration at a high level. Setup guides for each provider:

| Provider | Setup guide | |---|---| | GitHub | Connect GitHub → | | Jira | Connect Jira → | | Slack | Connect Slack → | | HubSpot | Connect HubSpot → | | Salesforce | Connect Salesforce → | | Aircall | Connect Aircall → | | RingCentral | Connect RingCentral → | | Zoom | Connect Zoom → | | Gong | Connect Gong → | | Gmail | Connect Gmail → | | MS Graph | Connect MS Graph → | | IMAP | Connect IMAP → | | BambooHR | Connect BambooHR → | | Kisi | Connect Kisi → | | Toggl | Connect Toggl → | | Lattice | Connect Lattice → | | 15Five | Connect 15Five → | | Culture Amp | Connect Culture Amp → | | Perdoo | Connect Perdoo → | | Weekdone | Connect Weekdone → | | Asana | Connect Asana → |

Org-wide vs. per-employee

Two install patterns:

  • Org-wide integrations — one admin connects, every employee benefits. Examples: GitHub Org App, Microsoft 365, Slack workspace install, Salesforce, BambooHR.
  • Per-employee integrations — each employee connects their own account via OAuth. Examples: personal GitHub, personal Gmail (when the org isn't on Workspace).

The integrations page has these split into two sections so you don't go hunting for the right card. You can also send invite-links to specific employees who haven't connected yet.

What each source contributes

Code & tickets

  • GitHub — pull requests opened / reviewed / merged, commits authored, comments left, time-to-merge. Highest-density source for engineering teams.
  • Jira — tickets opened / resolved / commented, sprint completion, ticket type breakdown. Good for tracking ticket throughput vs. the cadence.
  • Linear — coming soon.

Communication

  • Slack — messages sent, channels active in, help replies given (replies in #help-* channels), threading depth. Useful for surfacing teamwork-shaped contributions.
  • Microsoft Teams — message volume, meeting load, transcript availability for call quality. Same shape as Slack signals.
  • Gmail / Outlook (via Microsoft Graph) — emails sent / received counts only, no content. Used as a presence signal (vs. used as content for review). Domain-wide M365 install pulls every employee at once.

CRM & sales

  • Salesforce / HubSpot — deals won, opportunities created, pipeline movement, customer activity. Specifically useful for AE / SDR / CS roles.

Phone & calls

  • Aircall — total calls, talk time, call outcomes from disposition codes.

Call quality (opt-in)

If your team uses Aircall, Microsoft Teams, or Zoom and records customer-facing calls, you can opt in to call quality analysis — Claude reviews each transcript against a rubric you write and produces a per-call quality score plus notable moments. This is purely customer-facing calls; internal team meetings are never recorded or analyzed (Teams / Zoom adapters explicitly require at least one external attendee per call before analysis runs).

Opt-in is gated behind a consent attestation at /admin?tab=integrations → Call Quality card. Free-form rubric — write it however you want.

HRIS & directory

  • Microsoft 365 (Microsoft Graph) — employees, manager relationships, start dates. The recommended HRIS source for any team on M365.
  • BambooHR / 15Five / Lattice / Culture Amp — directory + manager hierarchy + (where applicable) goal data.

Goals

  • Asana / Perdoo / Weekdone / Lattice / 15Five / Culture Amp / BambooHR / Jira (with goal projects) — pulls active and completed goals for each employee in a cycle. Goals show up as a separate signal in the dashboard plus a [GOALS] block in the review prompt.

How signals reach a review

For each cycle, every connected source's data is pulled into raw_signals rows scoped to that cycle (each row has cycle_id, signal_date, source, event_type, and a JSON payload). When you generate a review, those rows are aggregated per source into the prompt's [GITHUB DATA], [SLACK DATA], etc. blocks alongside the structural context.

You don't need to think about this layer most of the time — the dashboard's signals strip is the user-facing view. But it matters for two things:

  • Cohort baselines — when generating a review, PerfCopilot computes the median value for the employee's team / department / org cohort for each headline metric. The AI is then instructed to ground claims comparatively ("above the team median for PR throughput") instead of absolutely.
  • Auditability — every signal is timestamped and traceable back to the source system. You can click any row in the signals strip to see which PR / message / ticket it came from.

Browse by category

If you'd rather read about a signal type instead of a specific provider:

What we don't do

A few things worth being explicit about:

  • We don't read the content of customer emails. Gmail and Outlook integrations pull metadata only — sender, recipient, timestamp, thread length. Subject lines and bodies are never sent to Claude or stored on our servers.
  • We don't record or analyze internal team meetings. The Teams and Zoom adapters require at least one external attendee per call. Internal-only meetings are dropped before any transcript leaves your provider.
  • We don't share your data with other PerfCopilot customers. Cohort baselines are computed within your org only.

See the Privacy and Terms pages for the formal version.

PreviousPeer feedbackNext 15Five