Catalog v1
Twelve launch-quality workflows across six categories. Every YAML public.
Three columns: shipped, in harness, researching. Live workflows from the catalog auto-feed the first column; forthcoming workflows auto-feed the second. No dates, no quarters — the cadence after October 2026 is daily, and the column is what the schedule is.
Live in the catalog or in production. Public YAML, harnessed runtime.
Twelve launch-quality workflows across six categories. Every YAML public.
Deterministic, in-browser, no account. Runs every workflow against scripted turns.
One URL per workflow, scope-locked, rotatable. Configs for Claude Desktop, ChatGPT, Cursor.
Step-level visibility for every run — confidence floors, escalation reasons, transitions.
Classifies an inbound message into one of six intents and routes to the matching downstream workflow.
Asks one targeted question when the inbound message is too vague to route, then re-routes with the answer.
Confirms whether an inbound message is a duplicate of an existing open ticket and merges or splits accordingly.
Collects structured refund-request information and produces a handoff packet for human review.
Walks the customer through a specific invoice line by line and answers questions grounded in the line items provided.
Handles cancellation intake, captures the reason, and confirms the effective date — without retention pressure.
Walks the customer through reproducing a bug and produces a clean, engineer-readable report.
Confirms whether a customer's report matches a known open issue and sets expectations on a fix.
Helps the customer find the existing feature that solves their problem, grounded in the product docs you supply.
Closes the AI conversation cleanly and produces the structured packet a human agent needs to pick up.
Confirms the issue is engineer-grade, collects the technical context, and writes the escalation note.
Asks the customer to confirm whether the issue is resolved and captures a one-line outcome.
Recognizes off-topic messages and declines politely without engaging with the off-topic content.
Authored, schema-checked, and queued for daily release after launch.
Pull the 90-day rolling history into the analytics stack your client already has.
Anthropic and OpenAI first. Pay your provider directly; retention halved when keys are yours.
Detects message language and routes to the matching localized workflow.
Assigns P1–P4 based on customer-stated urgency and message signals.
Routes flagged customer accounts directly to a senior queue with structured context.
Sends a single-turn acknowledgment with next-business-hour context when the queue is closed.
Walks customer through updating a failed payment method.
Confirms an upgrade or downgrade and sets expectations on prorated billing.
Captures the structured information required to transfer an account to a new owner.
Confirms what document the customer needs and routes to billing.
Answers pricing questions strictly from the published pricing page.
Step-by-step setup guidance grounded in your client's documentation.
Targeted intake for failures at the boundary of two systems.
Identifies where in onboarding the customer is stuck and produces a structured handoff.
Captures the scope and format the customer needs.
Collects browser, OS, and extension information for diagnostic handoff.
Structured collection of performance complaints with timestamps and scope.
Captures the reason for manager-level attention with structured grounding.
Routes incident-related messages into the active incident channel with context.
Recognizes legal-affecting language and stops AI handling immediately.
Detects press inquiries and routes to the communications inbox.
Recovers gracefully when prior context has been lost between turns.
Pauses and confirms with the customer when model confidence drops below threshold.
Collects a numeric satisfaction score with a single optional comment.
Produces the structured summary your client's archive expects.
Confirms a fix held N days after the reported resolution.
Closes a conversation gracefully when the customer has thanked the team.
Extracts a structured lesson from a notable ticket for the operator's records.
Picks up a reopened ticket cleanly with prior context as input.
Names the line, sets the consequence, hands off if it continues.
Recognizes signals of a minor user and escalates immediately.
Recognizes safety-critical language and follows a published protocol.
Captures the structured deletion request your client's privacy team needs.
Captures the structured access request with verification questions.
Inserts the required disclaimer in regulated industries before continuing.
Recognizes attempts to impersonate the operator or vendor.
Open questions worth answering. No commitment, no date — just on the page.
Let practitioners pin workflow versions in a config file and audit drift between runs.
Track YAML changes and surface breaking changes to subscribers before they ship.
Pipe run logs into the BI tool the client already runs (Snowflake, BigQuery, ClickHouse).
Single-region deployment for practices required to keep data inside the EU.
Let buyers preview the tuned variant against scripted turns before the commission ships.
Declarative chaining so a routing workflow can hand off into a downstream one with one config edit.
If something on the Researching column matters to your practice, email diego@agora.*with the shape of how you’d use it. Items move from Researching to In harness when someone takes the time to write that email.