New · MCP server · v1.9.5

Run your customer experience from inside Claude.

Connect ARIA to Claude Desktop in 60 seconds. Ask in natural language; get grounded, audited answers about your customers. Every call is row-level-security scoped, redaction-floor enforced, and written to your org_audit_log.

See the 6 Playbooks

Bearer-token auth today. SOC 2 Type 1 targeted Q3 2026 (currently in-progress) · GDPR · PIPEDA · EU AI Act Article 13 design-aligned.

What is MCP?

The open standard for AI assistants that can actually do things.

Model Context Protocol is the open standard adopted in 2025 for letting AI assistants call tools inside other apps. ARIA's MCP server lets Claude, ChatGPT, or any agent built on the OpenAI Agents SDK reach into your customer data — grounded, redacted, audited — without copy-paste.

See it in 30 seconds

Paste config · restart Claude · type a slash-command.

Demo recording in flight

Lands with v1.9.5 soft-launch — watch this slot

Honest hedge: the v1.9.5 page ships with a placeholder while the screen-only demo (no founder face, Pika-pattern) is recorded during the soft-launch window.

Why this is different

Not a chatbot widget. An AI-native customer-experience layer.

The category MCP creates is operating modes, not integrations. ARIA inside Claude is the same product surface as ARIA in your browser — same row-level security, same audit log, same Why? button on every output. The interface just moved.

Most "AI customer support"
ARIA inside Claude
Chatbot bolted onto a help desk
An operating mode for your customer experience
Plugin sells your data back to its model vendor
Bring-your-own-key — your prompts, your costs, your audit trail
Generic LLM with no memory of who Sara is
7-zone customer memory with row-level security on every read
No way to ask why it answered that
Per-decision explainability — same data your regulator will see
Three ways to connect

Same 21 tools. Same audit log. Whichever client you prefer.

Generate one key. Connect Claude Desktop, ChatGPT, or any agent built on the OpenAI Agents SDK. Switching clients does not change what ARIA can do — just which window the operator works in.

1. Generate your MCP key

Open Settings → Integrations → MCP in the ARIA app and click Generate key. Org-scoped, RLS-enforced, audit-logged.

2. Paste config & restart Claude

Open claude_desktop_config.json and paste the snippet below. Replace aria_mcp_xxx with your key.

{
  "mcpServers": {
    "aria": {
      "url": "https://mcp.aria.simplification.io/mcp",
      "headers": {
        "Authorization": "Bearer aria_mcp_xxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Quit and relaunch Claude Desktop. ARIA's 21 tools and 6 Playbooks appear in the slash-command menu.

Full walkthrough: docs · Connect Claude Desktop. Public docs site is in flight; the in-product user-guides are live for signed-in customers.

Generate your MCP key

One key, all three paths. Rotate or revoke any time at Settings → Integrations.

Anthropic Connectors directory: submission pending

Submission is paired with our v1.10 train. Once approved, one-click install in Claude Desktop replaces the JSON paste shown below. Until then, the manual paste is the canonical path — same key, same scope, same audit trail.

Playbooks

Six packaged Playbooks. Type a slash-command, get an answer.

Playbooks chain ARIA tools into operator-shaped briefings. They replace the 5-tab Monday-morning ritual with a single line typed into Claude.

Starter+

/aria:morning-brief

Morning brief

Daily attention brief — customers worth attention this morning, ordered by risk, with recommended next actions.

Starter+

/aria:churn-watch

Churn watch

Customers crossing churn-risk thresholds — with trajectory data and grounded citations to support intervention.

Agency / Enterprise

/aria:agency-rollup

Agency rollup

Cross-brand rollup — KPIs and outliers across every sub-brand under your parent org. Agency and Enterprise only.

Starter+

/aria:dsar-assist

DSAR assist

Drafts a DSAR export bundle for legal review — assembles the data set and flags redaction calls.

Starter+

/aria:bench-review

Bench review

Reviews recent agent decisions — low-confidence trajectories, grounding quality, and peer-review disagreement, ranked.

Growth+

/aria:compliance-snapshot

Compliance snapshot

Current compliance posture — controls in place, in progress, and planned. Honest hedges. Growth tier and up.

The DSAR-assist Playbook produces drafts for legal review — not ready-to-file documents. Compliance-snapshot is a posture report, not a third-party attestation. Honest hedges live at the bottom of every output.

Marketplace · Path 4

Discover battle-tested Playbooks from the ARIA community.

Operators publish their winning Playbooks to the marketplace. Install one in two clicks — they live alongside ARIA’s built-ins inside your MCP-connected Claude. ARIA reviews each for safety; outcomes are not endorsed.

support

Support Deflection Pack

Reduce repetitive ticket volume by 30%+ with five battle-tested prompt patterns for first-touch deflection, refund triage, and shipping-delay empathy.

by @@acme-co487 installs
4.7★ (92)
sales

Outbound Sales Cadence

A 7-touch outbound sequence that books meetings without spamming. Pulled from 14 months of Beta Corp's own data — open rates, reply rates, and meeting-booked rates per touch.

by @@beta-corp312 installs
4.5★ (67)
ops

Incident Postmortem Template

Run a blameless postmortem in 45 minutes. Outputs a five-section doc — timeline, impact, root cause, what worked, action items with owners and due dates.

by @@gamma-llc256 installs
4.6★ (48)

Marketplace listings are creator-published. ARIA reviews each one for safety but does not endorse outcomes from third-party Playbooks. Per-customer install limits apply: Solo:5 · Starter:20 · Growth+ unlimited.

Auth + rate limits

Predictable budget. Predictable scope.

Rate limit
100 req/min per key
Configurable on Enterprise.
Scope
Org-scoped
RLS enforced on every tool.
Rotation
Self-serve
Settings → Integrations.
Service-mode JWT
5-min TTL
For org-scoped automations.
OAuth 2.1 + PKCE
v1.9.5.12 LIVE
.well-known discovery doc.
Trust posture

Grounded · Redacted · Audited.

The same controls ARIA enforces in the browser run on every MCP call. Bearer-token auth today; OAuth ships in v1.9.5.6.

Grounded

Every answer cites the knowledge-base entries and customer records it pulled from. The Why? button surfaces grounding score, model used, latency, and cost.

Read more in Trust Center

Redacted

Privacy floor enforced before responses leave the API — PII tokenised, secrets scrubbed, sub-brand data stays in its lane via row-level security on every read.

Read more in Trust Center

Audited

Every MCP tool call writes an append-only org_audit_log entry. Same trail an EU AI Act Article 13 reviewer expects. SOC 2 Type 1 targeted Q3 2026.

Read more in Trust Center

Compliance posture honest hedge: ARIA is design-aligned with SOC 2 and ISO 27001 controls; neither has been third-party attested. The Type 1 audit is scheduled for Q3 2026. Drafts for legal review where compliance content surfaces (DSAR, data-residency, BAA).

Community

Talk to the team and other operators building on ARIA MCP.

Both spaces are public. Every release note, breaking change, and shipped tag is announced first in Discord; Slack is where customer- success and enterprise-buyer conversations land.

Pricing

Included on every paid tier. Read-only on Solo.

MCP is not a separate add-on. The same 21 tools and 6 Playbooks ship with every paid plan. Solo gets read-only access; Starter and above get read + write. Bring your own LLM key on Growth+.

v1.9.5 honest hedges, all in one place: SOC 2 Type 1 targeted Q3 2026 (currently in-progress) · ISO 27001 planned (not certified) · Bearer-token auth today, OAuth in v1.9.5.6 · Demo video records during the v1.9.5 soft-launch window · Drafts machine-translated for FR/AR (pro pass owed).