Features
Everything an agent needs to do real work.
Building, orchestration, observability and governance in one platform — with the controls operations, security and finance teams expect before anything goes live.
01
/
Build
Draft, test and ship agents without a sprint.
Describe the job in plain language and refine it in a visual builder, with a test run beside every change.
01
/
Build
Draft, test and ship agents without a sprint.
Describe the job in plain language and refine it in a visual builder, with a test run beside every change.
Visual agent builder
Lay out triggers, instructions, tools and conditions on one screen, with a test run beside every change.
Visual agent builder
Lay out triggers, instructions, tools and conditions on one screen, with a test run beside every change.
Plain-language instructions
Write the job like a runbook. Cairn turns it into steps you can review, version and diff.
Plain-language instructions
Write the job like a runbook. Cairn turns it into steps you can review, version and diff.
Tool library
120+ prebuilt tools with typed inputs, plus custom tools from any REST API or webhook.
Tool library
120+ prebuilt tools with typed inputs, plus custom tools from any REST API or webhook.
Knowledge sources
Ground answers in Notion, Google Drive, Confluence and help centers, with citations in every run.
Knowledge sources
Ground answers in Notion, Google Drive, Confluence and help centers, with citations in every run.
Templates
Start from production-tested workflows for support, sales, operations, finance and more.
Templates
Start from production-tested workflows for support, sales, operations, finance and more.
Test runs
Replay real tickets, emails or records against a draft agent before it touches production.
Test runs
Replay real tickets, emails or records against a draft agent before it touches production.
02
/
Run
Workflows that survive the real world.
Durable execution, branching and human-in-the-loop steps, so runs finish correctly even when the systems around them don’t.
02
/
Run
Workflows that survive the real world.
Durable execution, branching and human-in-the-loop steps, so runs finish correctly even when the systems around them don’t.
Triggers
Start runs from app events, webhooks, inbound email, forms or a cron schedule.
Triggers
Start runs from app events, webhooks, inbound email, forms or a cron schedule.
Branches and conditions
Route on amount, confidence, customer tier or anything the agent finds along the way.
Branches and conditions
Route on amount, confidence, customer tier or anything the agent finds along the way.
Multi-agent handoffs
Pass work between specialised agents with typed context and one continuous trace.
Multi-agent handoffs
Pass work between specialised agents with typed context and one continuous trace.
Human-in-the-loop steps
Pause for a decision in Slack, Teams or email, then resume exactly where the run left off.
Human-in-the-loop steps
Pause for a decision in Slack, Teams or email, then resume exactly where the run left off.
Retries and timeouts
Backoff, idempotency keys and fallbacks for every external call, so work never runs twice.
Retries and timeouts
Backoff, idempotency keys and fallbacks for every external call, so work never runs twice.
Schedules
Run digests, reconciliations and audits on a schedule, in your team’s time zone.
Schedules
Run digests, reconciliations and audits on a schedule, in your team’s time zone.
03
/
Observe
Know what every agent did, and why.
Each run leaves a trail you can search, replay and turn into tests — so quality improves with every change instead of drifting.
03
/
Observe
Know what every agent did, and why.
Each run leaves a trail you can search, replay and turn into tests — so quality improves with every change instead of drifting.
Run traces
Every input, tool call, decision and output, timestamped and searchable.
Run traces
Every input, tool call, decision and output, timestamped and searchable.
Eval suites
Turn real runs into test cases and score decisions, actions and tone on every change.
Eval suites
Turn real runs into test cases and score decisions, actions and tone on every change.
Replay
Re-run any trace against a new version to see exactly what would change.
Replay
Re-run any trace against a new version to see exactly what would change.
Analytics
Autonomy rate, resolution time, cost and time saved per agent, team and workflow.
Analytics
Autonomy rate, resolution time, cost and time saved per agent, team and workflow.
Alerts
Get notified in Slack, PagerDuty or email when error rates, costs or escalations spike.
Alerts
Get notified in Slack, PagerDuty or email when error rates, costs or escalations spike.
Cost controls
Budgets and rate limits per agent, with forecasts before you reach them.
Cost controls
Budgets and rate limits per agent, with forecasts before you reach them.
04
/
Govern
Autonomy on your terms.
Start in shadow mode, gate risky actions behind approvals and keep a complete record of who — or what — did what.
04
/
Govern
Autonomy on your terms.
Start in shadow mode, gate risky actions behind approvals and keep a complete record of who — or what — did what.
Shadow mode
Agents propose actions without taking them, so you can compare against your team first.
Shadow mode
Agents propose actions without taking them, so you can compare against your team first.
Approval gates
Require sign-off above thresholds you set — by amount, action type or confidence.
Approval gates
Require sign-off above thresholds you set — by amount, action type or confidence.
Scoped permissions
Grant each agent only the tools, records and actions it needs, enforced at runtime.
Scoped permissions
Grant each agent only the tools, records and actions it needs, enforced at runtime.
Audit log
An immutable record of every change, approval and action, exportable to your SIEM.
Audit log
An immutable record of every change, approval and action, exportable to your SIEM.
SSO and SCIM
SAML single sign-on and automated provisioning with Okta, Entra ID and Google.
SSO and SCIM
SAML single sign-on and automated provisioning with Okta, Entra ID and Google.
Data residency
Keep data and traces in the EU or US, with customer-managed keys on Enterprise.
Data residency
Keep data and traces in the EU or US, with customer-managed keys on Enterprise.
05 — DEVELOPERS
Everything in the interface is in the API.
Define agents as code, review them in pull requests and promote them through environments from CI. The builder and the API stay in sync both ways.
Versioned REST API with typed TypeScript SDK
Webhooks for every run, step and approval
CLI for local testing and CI promotion
Sandboxed custom tools in TypeScript or Python
create-agent.ts
import { Cairn } from "@cairn/sdk"; const cairn = new Cairn({ apiKey: process.env.CAIRN_API_KEY }); const agent = await cairn.agents.create({ name: "Refund Resolver", trigger: { app: "zendesk", on: "ticket.created" }, instructions: "Refund damaged items up to $150.", tools: [ "shopify.orders.read", "stripe.refunds.create", "zendesk.tickets.reply", ], approvals: [{ when: "amount > 150", in: "#finance-ops" }], mode: "shadow", }); console.log(agent.id); // agt_8Hq2Lm4x
See it on your work
Bring one workflow. We’ll build the agent live.
In 30 minutes an engineer maps a real process to an agent, runs it in shadow mode on sample data and shows you the trace.
Free 14-day trial · No credit card · SOC 2 Type II
See it on your work
Bring one workflow. We’ll build the agent live.
In 30 minutes an engineer maps a real process to an agent, runs it in shadow mode on sample data and shows you the trace.
Free 14-day trial · No credit card · SOC 2 Type II