Reply Jason AI

Agents take authenticated action across several classes and the catalog is open rather than fixed.

Named: two-way CRM sync with Salesforce, HubSpot and Pipedrive; sending mailboxes connected by Gmail and Office 365 OAuth with IMAP and SMTP support; LinkedIn accounts with daily action limits and connection monitoring; Google Calendar for booking; a Chrome extension; and a published integrations directory.

The way to add your own is unusually strong: the full REST catalog, a live MCP server and a CLI expose contacts, sequences, conversations, tasks, templates, mailboxes and suppression rules as callable modules, so anything not pre-integrated can be wired by the customer.

Sourcereply.io/integrations, agents.reply.io API catalog, docs.reply.io api-referenceread 2026-09-01

The platform carries the control flow. Multichannel conditional sequences mix email, LinkedIn and call steps in one campaign with conditions that adapt to prospect behavior, and the vendor gives that its own product page.

Playbooks hold the strategy the agent executes, either authored by the customer or generated by Jason, and re-engagement rules define when and how to return to prospects who went quiet. Call steps are inserted where most effective and generate a per-prospect script as a task.

On the API side sequences are a first-class object that can be created, populated and controlled, background jobs track long-running bulk operations, and webhook events can trigger the next stage.

Sourcereply.io/jason-ai-sdr-multichannel-conditional-sequences, reply.io/jason-ai automation section, agents.reply.io Sequence and Webhooks APIsread 2026-09-01

The customer points the agent at their own material and it is retrieved when the agent writes.

A knowledge base holds company information, product details and objection handling, and the vendor states Jason uses it for smart replies and to stay on track; offers define what is sold to each customer type with pain points, solutions and proof such as case studies; playbooks hold templates, instructions and custom prompts.

All three are callable through the AI SDR API, so the corpus is maintained rather than assembled per request. Personalization additionally researches each prospect across LinkedIn, company websites and other public sources, with the customer selecting which personalization points are used.

Sourcereply.io/jason-ai strategy setup and personalization sections, agents.reply.io AI SDR API, reply.io/blog AI SDR knowledge base release postread 2026-09-01

Oversight is configurable and the vendor gives it its own product page. Approval mode lets the customer review and approve every message before it goes out with full context shown, and reply handling is switchable between automatic mode, where the agent answers on the customer's behalf, and approval mode, where it drafts and waits. The approval queue is callable through the AI SDR API, so the gate is programmable rather than only a UI toggle.

Guardrails sit alongside it: suppression and blacklist rules by domain, email and exception are enforced platform-wide as do-not-contact compliance, LinkedIn accounts carry daily action limits, and the MCP toolset annotates every tool as read-only or destructive so a client can auto-gate risky calls, with confirmation gates and dry-runs in the CLI and skills.

Sourcereply.io/jason-ai-sdr-approval-mode, reply.io/jason-ai reply handling section, agents.reply.io AI SDR and Suppression APIsread 2026-09-01

A readable trust page exists and almost nothing on it is specific. It describes access controls, authentication measures, system monitoring, vulnerability management, backup and recovery, incident response with notification, encrypted transport, and no selling or training on customer data, and it makes a Data Processing Agreement and an Acceptable Use Policy available.

Scoped API keys with granular permissions such as contacts:read and sequences:operate are a real access model for the API surface. No certification is named (no SOC 2, no ISO 27001, no audit firm and no report), additional security materials are offered only on request or under NDA, the sub-processor list is available on request rather than published, and no SSO or SAML for the application, role model or customer-facing audit log is documented.

Sourcereply.io/trust-page, agents.reply.io scoped key documentationread 2026-09-01

Run-level visibility reaches the why, not just the how many. Source transparency is documented as a first-class feature: the customer can review every source the agent used to craft a message, with direct links provided so the reasoning can be verified. Around it the Analytics API exposes activity feeds, email, LinkedIn, call and task overviews, per-sequence statistics, channel efficiency, team performance and the meetings list, background-job tracking covers bulk operations, and mailbox and LinkedIn account health are monitored continuously.

Sourcereply.io/jason-ai personalization section, agents.reply.io Analytics and Platform Utilities APIsread 2026-09-01

State persists across a long-running motion: the unified inbox holds threads across email and LinkedIn that the agent reads, classifies and responds to over weeks, re-engagement rules govern returning to prospects who stopped answering, and contact records with AI custom fields form a CRM-grade store the agent works against. No memory layer with a stated scope and lifetime is documented; what persists beyond the conversations is the knowledge base and the authored playbooks and offers, and nothing describes the agent accumulating learning across runs or writing to a memory it later reads back.

Sourceagents.reply.io Conversations and Contact Data APIs, reply.io/jason-ai reply handling sectionread 2026-09-01

Cloud-only with no customer control over data location. The trust page states data may be processed in the United States and in other countries where the vendor or its authorized service providers operate, and offers no region selection, no residency commitment naming regions, and no single-tenant, private-cloud or self-hosted option.

What covers the transfers is legal instrument rather than residency: Standard Contractual Clauses, the UK International Data Transfer Addendum, and the EU-US, UK Extension and Swiss-US Data Privacy Frameworks, with a Data Processing Agreement available. No deployment option beyond shared multi-tenant software as a service appears anywhere on the site or in the plan descriptions.

Sourcereply.io/trust-pageread 2026-09-01

Packaged assets ship in several browsable forms. Playbooks are prebuilt outreach strategies the customer adopts or has Jason generate, with campaign strategy setup marketed as launching proven strategies in minutes. Reusable email and sequence templates with variable rendering are a first-class object with their own API.

For agents, eighteen installable skills ship in three packs covering planning, audience, launch, inbox, analysis and guardrails, installed by a single command. Eight named workflow recipes are published and browsable, each aimed at a named job: find and contact prospects, inbound lead follow-up, autonomous AI SDR, list hygiene, A/B testing, account-based campaigns, LinkedIn-first outreach and re-engaging dormant prospects.

Sourceagents.reply.io skills and workflows pages, reply.io/jason-ai-sdr-strategy-setupread 2026-09-01

The agent starts on signals rather than on a person typing, and the signal set is named: hiring activity, technology usage, company growth, LinkedIn engagement, website visits and competitor following, with intent signals given their own product page. Webhooks push every reply, open, click, bounce and LinkedIn event to a customer endpoint, so external systems can trigger the next action rather than polling.

Sending schedules and holiday calendars are configurable through the API. Channel coverage runs across email from the customer's own mailboxes, LinkedIn connection requests, messages, InMails and AI voice notes, phone call steps, and a unified inbox, with a mobile app pushing notifications the moment a prospect responds.

Sourcereply.io/jason-ai-sdr-intent-signals, reply.io/jason-ai targeting and engagement sections, agents.reply.io Webhooks and Events APIread 2026-09-01

Model choice is an explicit product surface with a named provider list. The Jason strategy setup documents AI model selection as a configuration step, stating the customer picks their AI engine from Claude, Gemini, Mistral or OpenAI, and the vendor's own release post announces model selection as a shipped feature alongside approval mode. The trust page corroborates that AI features run on authorized third-party providers and that customer data is not used to train generalized models for other customers.

Sourcereply.io/jason-ai strategy setup section and reply.io/blog smarter-outreach release postread 2026-09-01

Four integration surfaces are published over one platform and one scoped key: a REST API at api.reply.io/v3 with OpenAPI and problem+json errors, a live MCP server at mcp.reply.io presented as a full toolset for Claude, Cursor and any MCP client, a CLI installable by npm as reply-cli, and eighteen installable agent skills in three packs.

Scoped permissions are granular, with contacts:read and sequences:operate given as examples, and the MCP surface carries machine-readable readOnlyHint and destructiveHint annotations, fail-fast schemas and stable error codes. Webhooks push every reply, open, click, bounce and LinkedIn event outbound. The vendor states plainly that everything documented is callable today unless labeled Coming soon or Concept, and MCP is served rather than merely consumed.

Sourceagents.reply.io API catalog and MCP page, docs.reply.io api-referenceread 2026-09-01

Two criteria are applied before and after a change ships. Before a message goes out, triple-verification, a named platform capability, validates it to protect sender reputation, a gate in the release path with a pass condition rather than an inspection.

Once a change ships, A/B testing of message variants is documented on the email automation surface and as a supported workflow, running variants against live traffic, killing the losers and promoting winners into reusable templates, which is a scored comparison producing a readable result. Those tests measure results on live traffic rather than before a message goes out.

Sourcereply.io/jason-ai-validation, reply.io/jason-ai multichannel section, agents.reply.io A/B test workflowread 2026-09-01

LinkedIn automation has its own product page and the documented actions are specific: connection requests, messages, follow-ups and AI-generated voice notes in the customer's own voice, with a LinkedIn Accounts API for connecting accounts and managing daily action limits, so the agent acts inside an interface the vendor does not control. No first-party page states how those actions are executed, so the mechanism is not documented.

Sourcereply.io/linkedin-automation, reply.io/jason-ai engagement sectionread 2026-09-01