What does Watari install in my Intercom workspace?

A single OAuth integration with conversation and conversation-part webhook subscriptions. No Messenger app surface, no Inbox app surface, no teammate training. You can scope which conversations reach Watari by Intercom team assignment at install time.

What Intercom data does Watari access, and what does it store?

Watari reads conversation parts (customer and teammate messages), conversation attributes (state, tags, custom attributes, team assignment), contact display name and email, and attachments (including screenshots) for the conversations you scope to the integration. Tokens and webhook secrets are AES-256-GCM encrypted before persistence. Customer conversations are processed for bug extraction and code mapping; they are not used to train any model.

How is this different from an Intercom-to-Jira integration?

A sync app copies a conversation into Jira as another ticket on another board. Watari turns the conversation into a draft pull request mapped to a real file and function in your repository — and writes the customer back when the fix is in production. You get a diff your reviewer can take over, not a duplicate of the original conversation.

You pay per Mapped Bug — a structured bug paired with the file and function in your repo, both at ≥ 0.7 confidence. PR drafts, customer-facing RCA, and Jira / GitHub Issues sync are bundled. The meter fires at persist time of a qualifying bug; non-bug conversations (sales questions, onboarding help, billing) never cross the gate. If we map to the wrong file, flag it as Mismapped within 7 days for a credit on your next invoice.

Does Watari work across multiple repositories?

Yes — Watari fans out across every repository scoring ≥ 0.7 relevance. Each contributing repo gets its own draft PR routed via that repo’s CODEOWNERS. The customer-facing RCA on the Intercom conversation is held until every contributing PR merges and the production deploy confirms the fix is live.

What if Watari can’t fix the bug?

Some bugs are infrastructure-only, content-only, or out-of-scope for the indexed code. Watari surfaces those with a not-fixable classification and a one-line explanation. The Mapped Bug is still useful (it carries structured repro, severity, and customer impact) — the PR draft step is the part that gets skipped.

Is my source code safe?

Tokens, refresh tokens, and webhook secrets are AES-256-GCM encrypted before they hit the database. Tenant data is isolated via Postgres row-level security on every table. We don't store your source code — only vector embeddings and metadata; the actual code is fetched from GitHub on demand and discarded after use. We never train models on your code. Full security and privacy details at watari.ai/security.

Intercom App Store · Customer Support

Watari for Intercom

Customer conversations become draft pull requests, mapped to the line.

Watari connects to Intercom via OAuth, extracts a structured bug from every qualifying conversation, and maps it to the file and function in your GitHub repository. The draft PR is opened with the patch and the test. The customer-facing RCA waits for your production deploy to confirm the fix is live.

Install Watari free See pricing

Trial: 14 days · 10 Mapped Bugs
Card required: No
Setup: OAuth + webhooks
Mismapped credit: 7 days

What is Watari for Intercom?

Watari for Intercom is an App Store listing that turns customer conversations into draft GitHub pull requests. It subscribes to Intercom conversation webhooks, extracts a structured bug, maps it to the responsible file and function in your repository at ≥ 0.7 confidence, and opens a draft PR routed via CODEOWNERS. The customer-facing RCA is posted back to the conversation only after the merge lands in a successful production deploy.

Why support and engineering teams install Watari for Intercom

Mapped to the file and function

Every qualifying conversation clears a dual ≥ 0.7 confidence gate (extraction and code-mapping) before it becomes a Mapped Bug — paired with the exact file and function in your repository, not a vague tag.

Vision-aware on screenshots

Customer screenshots, video frames, and image attachments are parsed by a vision-aware model alongside the conversation transcript. Repro steps and severity are extracted automatically.

Team-scoped intake

Configure a single team filter at install — only conversations assigned to that team reach Watari. Other Inbox queues (sales, billing, onboarding) stay untouched. Spam, not_a_bug, and watari-skip tags also short-circuit extraction.

Draft pull request in your repo

Watari writes the patch and the test, runs them through type and lint checks, and opens the draft PR routed via CODEOWNERS — type-clean and ready for review.

Customer reply gated on deploy

The customer-facing RCA is held until the GitHub deployment_status webhook fires success and the Compare API confirms the merge SHA is live in production. A later rollback parks the RCA and surfaces a banner.

Multi-lingual, multi-channel dedup

Customer verbatim is preserved in the original language. Duplicate reports across Intercom + Zendesk + email cluster into one canonical bug — one PR, RCA fan-out to every linked conversation, idempotent on Intercom’s API so retries never double-post.

How Watari turns an Intercom conversation into a draft PR

Install Watari from the Intercom App Store
One-click install from the Intercom App Store listing. Watari authenticates via OAuth — no redirect URI configuration required on your side. No Messenger app, no Inbox app, no agent training: Watari runs entirely on Intercom webhooks.
Watari subscribes to conversation webhooks
New conversation events and conversation-part (reply) updates are delivered to Watari via Intercom webhooks. Existing teammate workflows are untouched. You can scope which conversations reach Watari by Intercom team assignment — only conversations assigned to the configured team trigger extraction.
Bug extracted into a structured, schema-validated record
Severity, reproduction steps, expected vs actual behavior, customer impact, and original-language verbatim — extracted by a vision-aware model with structured outputs. Below 0.7 extraction confidence, the bug is queued for review and never billed.
Bug mapped to the file and function in your repository
Watari searches your indexed code and narrows the candidates to the exact file and function responsible, each with a 0–1 mapping confidence. The Mapped Bug requires both extraction and mapping at ≥ 0.7 — below that, the bug queues for review and is never billed.
Draft PR opened, CI watched, customer reply held for deploy
The draft pull request is opened in GitHub with the patch, the test, and reviewers from CODEOWNERS. CI status is observed; reviewer comments trigger non-force regenerations. After the merge lands in a successful production deploy, the customer-facing RCA is posted back to the original Intercom conversation in the customer’s language.

What Intercom data Watari accesses

Watari accesses only the Intercom data required to extract a structured bug and map it to code. All scopes are requested at install time via OAuth; you control which conversations reach Watari by scoping to a specific Intercom team at setup.

Read: Conversation parts (customer and teammate messages), conversation attributes (state, tags, custom attributes, team assignment), contact display name and email, and attachments (including screenshots) for conversations you scope to the integration.
Write: A single customer-facing reply posted back to the conversation after the fix is live in production (the RCA). Optional: an internal note linking to the draft PR.
Store: Extracted bug records, code-mapping results, and a token reference. Tokens and webhook secrets are AES-256-GCM encrypted before persistence. Tenant data is isolated via Postgres row-level security on every table.
Never used for: Model training. Your conversations and source code are not used to train any model — ours or any third party’s. Full detail at watari.ai/security.

Pricing

You only pay when Watari delivers a Mapped Bug — a structured bug paired with the file and function responsible, both at ≥ 0.7 confidence. PR drafts, customer-facing RCA, and Jira / GitHub Issues sync are bundled.

Starter

$400/mo

20 Mapped Bugs included · $25/bug on-demand

For small support teams routing their first Mapped Bugs.

Growth

$900/mo

60 Mapped Bugs included · $20/bug on-demand

For teams running customer support at production volume.

Scale

$1,800/mo

150 Mapped Bugs included · $15/bug on-demand

For multi-repo SaaS with a dedicated support function.

Indian buyers (with GSTIN) are billed in INR; everyone else in USD. 14-day free trial · 10 Mapped Bugs included · no credit card required. Annual billing saves 17%. Full plan comparison at watari.ai/pricing.

Security & privacy

OAuth-only install — no shared API keys.
Tokens and webhook secrets AES-256-GCM encrypted at rest.
Postgres row-level security on every tenant-scoped table.
Your conversations and code never train any model.
HMAC-verified inbound webhooks on every request.
Per-region data residency on request (details).

Frequently asked questions

Your next support ticket arrives as a draft PR.

Connect Zendesk or Intercom, install the GitHub App. Tickets land mapped to the file, function, and line — ready for your reviewer to take over.

Start free trial

Trial length: 14 days
Bugs included: 10 Mapped
Card required: No
Mismapped credit: 7 days
Cancel: Any time

You only pay when we know what to change.