Shallow
A Slack bot is not enough
Real integration reads context, takes action, and stays accountable in the tool.

Integrations
Native connectors and MCP, with a signed record on every action.
The category
Integrations should do more than post notifications. Each action stays scoped, OAuth-bound, and logged in the workspace with the same provenance as a generate run.
Least-privilege OAuth per action class. No `repo:all` for a single PR.
Two-way sync with Jira, Linear, GitHub, Slack, Figma, and Google Workspace.
MCP servers treated as native tools with the same approval gates.
Every external action carries a provenance ID in the workspace.
One “new” affordance and a project pipeline that mirror Slack, Jira, GitHub, Figma, Linear, and Notion — with OAuth-bound actions and provenance on every writeback.


The gap
Most AI tools ignore your stack or ship a shallow Slack bot. Both make your workflow worse.
Shallow
Real integration reads context, takes action, and stays accountable in the tool.
Glue
Zapier, cron, and webhooks become maintenance your team did not plan for.
Risk
Broad `repo:all` asks kill the integration in review.
Silos
Brief, ticket, code, and deploy live in different places; AI sees one.
Pick the integration. Authorize once. Same context and provenance everywhere.
Thread to ticket. Channel standup digest. @-mention for PR review with context blocks loaded.
Stays in the workspace with plan and checks
Two-way sync. Tickets created either side keep context and provenance.
Stays in the workspace with plan and checks
Read-only first, write on approval. PRs, reviews, branches, status checks signed by the approver.
Stays in the workspace with plan and checks
Official API. Frame URLs become typed components; design tokens detected.
Stays in the workspace with plan and checks
Deploy, monitor, roll back. Every deploy carries a provenance bundle.
Stays in the workspace with plan and checks
Docs round-trip for PRDs, Sheets for experiment results, Calendar for meeting linkage.
Stays in the workspace with plan and checks
Ingest and writeback so your wiki stays a context source as code changes.
Stays in the workspace with plan and checks
Bring any MCP server with the same approval, provenance, and context guardrails.
Stays in the workspace with plan and checks
Arvad sits where the thread keeps the plan, staging, spend, and checks.
| Capability | Most AI tools | Arvad integrations |
|---|---|---|
| Depth | Notifications | Read, reason, and write |
| Context across tools | Per tool | One workspace context |
| OAuth scopes | Broad | Least-privilege per action |
| Token rotation | You manage it | Automatic |
| Provenance on actions | No | Every action signed |
| Approval gates | No | Per integration, per action class |
| MCP | Not supported | First-class |
| Audit log | Maybe | Tamper-evident, queryable |
Pick an integration, authorize with least-privilege scopes, set approval gates. Every action signed and logged.
Pick an integration, authorize with least-privilege scopes, set approval gates. Every action signed and logged.
Least-priv
Least-privilege scopes per action
No `repo:all` when one repo would do.
Auto
Automatic token rotation
Expired tokens fail closed.
SSO
SSO-bound identity
Off-boarding revokes that user’s Arvad access.
BYOC
Bring your own cloud
Available when your deployment model supports VPC hosting.
Limits stated plainly. No demo theater.
One question
If you cannot, you still have a chat log. Arvad keeps that work in one workspace.
Keep exploring
Scoped OAuth, MCP, and provenance on every surface.