Your AI Workforce is Here! Agents are like worker bees that perform numerous digital jobs behind-the-scenes for you!
Built for founders and solo operators who want inbox and calendar work off their plate.
Get a daily brief, meeting prep, and reply drafts that run on your schedule—with approvals when it matters. Not just chat.
Inbox triage | Reply drafts | Calendar prep | Meeting follow-ups | Lead outreach | Website Q&A | Research briefs | Customer support | Social drafts | KPI summaries | Admin automation
An agent is a worker bee you configure once—then it runs in chat, on your website, on a schedule, or from webhooks & email. Start with inbox + calendar, then add weekly post drafts with the same guardrails. Eval baselines keep autopilot safe. No coding required.
Start free Free tier · No credit card to start
Recording a walkthrough (~3 min)? How-to script & storyboard · Recording checklist
After recording: Host & link your video · Funnel SQL: Founder funnel queries
About ten minutes: save an agent, test it with a realistic prompt, put something on autopilot, then save a performance snapshot so future updates do not quietly break what works.
Returning user? Log in or sign up for cloud sync, saved agents, and background autopilot.
Founder value: inbox and calendar workflows with draft-before-send, one place for embed · schedule · webhook · API, and eval baselines so upgrades do not silently break what you ship. Story and Ops packs give you logo, landing, pitch, email, social, and admin agents on the same rails.
New here? Start with the full founder stack, send one test in Test chat, open Autopilot under Go live (schedule · embed · webhooks · email-in), then set an eval baseline in Model & settings before you rely on the agent daily (Founder path).
Pick a template and click Use as agent to create a saved agent in one click. Each card’s Autopilot line shows how well it runs in the background via schedule, webhook/API, or site embed (see Go live → Autopilot).
Describe the repeated task in one sentence and we will recommend a template with safe defaults.
Try inbox, social, support, website monitoring, scheduling, or pitch outreach.
Why this lane: we match your intent to the most reliable background trigger for that workflow.
Loading proven agents from the public directory…
Agents others have shared. Filter by category or search, then clone to add a copy.
No public agents match. Change filters or click Refresh.
Or build from scratch: name your agent, fill in the basics below, then click Generate Agent.
We'll prefill the goal and suggest a name. Then tweak the form below.
Save your agent here, then pick it from the list below to test it or share it.
Select a cartoon or realistic portrait; it fills the preview square. Without an image, initials use default colors.
Your agent name shows here
Pick an avatar to apply instantly. No generation cost.
Sign in, upload optional reference image, tune framing (zoom/pan), then generate options.
Save your agent here, then pick it from the list below to test it or share it.
Agents you’ve saved or that were shared with you. Click Activate Agent to load one. Your role (Owner, Editor, Runner, or Viewer) controls publish, edits, and runs—see the banner above when a shared agent is active.
No agents yet. Create your Founder Story pack or Founder Ops pack above (or use a template and click Use as agent).
Same list as above, with status (draft/live) and quick actions. Click Refresh to see updates.
No agents to monitor yet. Save an agent above to see it here.
Type a message and click Send. If you haven’t saved the agent yet, use Simulation—it works right away. After saving, Auto or API will use your live agent.
Simulation = instant replies for testing. API = real agent run (requires a saved agent).
Active voice: --
Premium voice requires paid tier and a configured server endpoint.
One topic per thread. Scroll when the list is long; hover a row to see the full title.
Thread memory: none yet.
Logos, signs, ads, brochures, cards, posters, invites, and more — draft visuals appear here. Review before print or publish.
One-click inserts into the message box (Jordan and other agents with saved snippets). Edit before Send.
Long pastes stay verbatim: spellcheck and autocorrect are off so browsers don’t “fix” names, locations, or production terms.
Images: Paid plans — up to 2 uploads per UTC day (max 2 per message; JPEG, PNG, WebP, GIF; about 4MB each). Works with agents that use tools (email, calendar, Slack, etc.). Images go to the model like text; do not upload secrets, IDs you would not paste into chat, or confidential screens. Use API or Auto test mode (not simulation-only). The line above shows remaining quota when you are signed in on a paid plan.
Sent as context.structuredData for this send only—so the model can reason over rows or objects alongside your message.
.csv or paste comma-separated text.curl), an export from a developer tool, or any { "key": "value" } / array you use in your stack.Tip: Trim to the columns or fields you care about so the agent stays focused. Respect the character limit and your own privacy—don’t paste secrets or full customer PII unless you intend to.
Reply text appears as it’s generated. With tools, the final answer streams when ready.
⌘↵ or Ctrl+↵ to send
Save the last reply so this agent can use it in future runs.
One path: Test chat → Integrations (Gmail/Calendar when needed) → Go live (embed, schedule, or webhook) → Eval baseline in Model & settings → optional Require baseline to publish for stricter ship gates.
Load a saved agent to see your checklist. Use Autopilot below so this agent runs without you opening the app—then tighten trust with eval baselines in Model & settings.
Founder launch path: 1) Logo & identity (Brand Bee) → 2) Landing & one-pager (Nova) → 3) Pitch & investor emails (Ivy) → 4) Connect Gmail/Calendar for Email & Calendar Butler → 5) Ops & social (Robin, Buzz, Jamie). Use the pack buttons above for Story Pack + Ops Pack in one click, then the trust path before Autopilot.
Guide: docs/FOUNDER-PATH.md · Trust & CI:
PRODUCTION-TRUST-LOOP.md · Example workflow: .github/workflows/agent-eval-example.yml
Where to use this agent: Embed · Slack · API · Email Butler setup · Calendar Butler setup
Pick how this worker bee stays busy on its own. We optimize for trust rails (approval-first side effects), durable progress (knowledge, schedules, memory), verification (eval suite + baselines), and harness defaults on founder paths—tighten any gap in Long-run posture below before you fully delegate.
Pick a lane to start.
Applies a recommended lane, turns approval mode on, checks trust guardrails, and publishes only when safe.
Load an agent, then click Go live safely.
Trust: Model & settings → require approval for risky tools · Founder autonomy guide
One file for trust reviews (no API keys).
Load a saved agent to see a short path from first test run to first live surface (schedule, publish, or webhook).
Load a saved agent to see trust status.
Applies safe defaults and suggests any remaining optional setup.
Warnings are suggestions; block items mean the deployment cannot run agents until fixed (e.g. missing server API key).
We will quickly verify trust checks before first publish.
Publish (if needed), generate your deploy kit, and copy links/snippets for website + app + API in one place.
Load a saved agent, then click Generate deploy & share kit.
Public hosted chat URL for this agent.
Opens this agent in FreeAgentMaker for your team/account context.
Use this endpoint to run the agent from backend or automation tools.
Copy/paste this HTML snippet into your website.
Starter copy for Slack, email, or social launch announcements.
Which AI powers this agent. Default is usually fine; change if you need more speed or quality.
0.3Lower = more consistent, higher = more creative (0–2).
Cap response length. Leave blank for server default.
Max runs per minute for this agent. Leave blank for no per-agent limit.
When enabled, those actions create pending items in the Dashboard instead of running immediately. Approve or dismiss each one.
When checked, POST /api/agents/:id/publish returns 403 until a baseline exists, unless the client sends
forcePublish. Pair with Test suite & evaluation below.
Create a copy to try changes without affecting the original.
First message visitors see in the chat widget on your site.
We send a POST with run result (runId, status, reply snippet) when a run completes. HTTPS only.
We POST here only when a run errors (rate limit, timeout, etc.). Cooldown: once per 15 min per agent. HTTPS only.
Or add multiple sub-agents below for orchestration (researcher, writer, etc.).
Call run with planOnly: true to get a draft plan, then call with plan: "…" to run with that plan in context.
Link to an OpenAPI 3.x spec. Add your API URL here; when supported, operations will be available as tools. You can also add custom webhook tools below.
Connect MCP servers in Integrations. Then choose which tools this agent is allowed to use. (Human-in-the-loop still applies when required.)
Define tools that call your HTTP endpoint. Use a snake_case name (e.g. lookup_user). The agent can invoke them automatically; no need to add the name to Extra tools.
Blocklist rejects runs containing those keywords. PII redaction replaces them in replies.
If set, Gmail/Outlook sends only go to addresses on these domains (comma-separated). Subdomains: use *.parent.com. Empty = no extra restriction.
Production trust loop: docs/PRODUCTION-TRUST-LOOP.md · Eval / CI runbook
Load a saved agent to see eval baseline status.
Add sample inputs and optional expected text. Run evaluation to run all tests and see pass/fail, latency, and reply per case. Advanced: a test case may include expectedTools (JSON array of substrings) — each must appear in this agent’s configured tool names.
Founder golden path: one-click starter cases (suite founder-golden-path) for instruction-following and safe-send posture — filter
Run suite to that name or run All. Default eval is chat-only (fast). Check Include live production path to run liveRun cases through the same tool path as real runs (approval forced on; slower).
Using Slack? Click Add to Slack (start here), then open Slack setup steps below.
SLACK_DEFAULT_AGENT_ID to this agent’s ID so the bot uses this agent. Invite the bot to a channel to start.Trust: Scope the bot to channels that should trigger runs. Side-effect tools respect approval mode and your blueprint; review production trust before wide rollout.
Quick Zoom onboarding (5 minutes): connect Zoom once, then your agent can create meeting links during runs.
Client ID + Client Secret.https://freeagentmaker.ai/api/integrations/zoom/callback and add the same value in allow list.meeting:read:list_meetings and meeting:write:meeting.ZOOM_CLIENT_ID, ZOOM_CLIENT_SECRET, ZOOM_REDIRECT_URI.
Callback URL helper: https://freeagentmaker.ai/api/integrations/zoom/callback
Trust: keep approval mode on for side-effect actions and reconnect Zoom immediately if tokens expire.
Simple tutorial (5-10 minutes): use this once per deployment, then your agent can receive and reply to WhatsApp messages.
Phone number ID and Access token from API Setup.WHATSAPP_ACCESS_TOKEN, WHATSAPP_PHONE_NUMBER_ID, WHATSAPP_VERIFY_TOKEN, WHATSAPP_APP_SECRET, and WHATSAPP_DEFAULT_AGENT_ID (or WHATSAPP_AGENT_ROUTES).https://freeagentmaker.ai/api/whatsapp/webhook and the same verify token value from WHATSAPP_VERIFY_TOKEN, then click verify/save.messages webhook field, redeploy your server, and send a test message from WhatsApp.
Optional routing JSON example: {"default":"your-agent-id"}
WHATSAPP_VERIFY_TOKEN once on the server to match Meta’s verify token, and set WHATSAPP_APP_SECRET so inbound POSTs are verified via X-Hub-Signature-256.WHATSAPP_ACCESS_TOKEN, WHATSAPP_PHONE_NUMBER_ID, and WHATSAPP_DEFAULT_AGENT_ID (or WHATSAPP_AGENT_ROUTES JSON for per-sender routing) once on the server. Inbound messages are deduped by Meta message id; short thread memory is kept per sender. Replies go back through WhatsApp Cloud.
Trust: WhatsApp is a high‑reach channel — keep approval mode on for side effects, and treat unknown senders like untrusted input.
Draft-first workflow (safe): generate newsletter drafts here, review/edit in Substack, then publish manually.
https://yourpublication.substack.com).Use Run agent as an action (e.g. Form → FreeAgentMaker → Gmail) or New run completed as a trigger (run finishes → next Zap step). Open Zapier goes to Webhooks by Zapier so you can build Zaps with your run URL (a dedicated Zapier app listing may be added later).
Trust: Treat your run URL and API key like passwords. Build with test Zaps first; production Zaps should use least-privilege keys and approval mode for side effects.
Action — Run agent: In Zapier add step Webhooks by Zapier → POST. URL: your run URL (copy above). Header: X-API-Key = your API key (create in Model & settings → API keys). Body: { "input": "your message" }. Use the reply from the response in the next step.
Trigger — New run completed: Set Notify URL when a run finishes to a Zapier Catch Hook. Each completed run POSTs JSON there (webhookProtocolVersion: 2 and related fields). Webhook v2 & Zapier examples.
These starters include concrete endpoint + payload examples for the top founder workflows (Phase 1-3).
Your app can send a message to the URL below and get the agent’s reply. Example with curl:
Trust: Rotate API keys if they leak. Successful JSON responses include run_receipt (short summary of the run); the X-FAM-Run-Receipt header mirrors it for logs. Prefer drafts and approval for email/calendar until you are confident.
Or use the endpoint URL in your app; load a saved agent to see it.
Trigger this agent from a Make scenario or n8n workflow: add an HTTP / Webhook module that POSTs to your run URL with the body below.
Make: use "Make a request" module — Method POST, URL below, Headers: Authorization: Bearer YOUR_TOKEN, Content-Type: application/json, Body = Raw JSON. n8n: HTTP Request node — Method POST, URL below, Authentication = Header Auth (Bearer YOUR_TOKEN), Body = JSON.
Receive emails and run this agent automatically. Copy the webhook URL and secret below, then configure your provider (SendGrid Inbound Parse, Mailgun Routes, etc.) to POST incoming mail to that URL.
Draft only keeps replies as suggestions you can copy or forward. Auto-send replies directly to the sender for matching emails.
X-Inbound-Secret with your secret. In Mailgun: Routes → forward to your webhook URL and add the same header. Your provider must send a JSON or form body with from, to, subject, and text (or we accept common variants).agentId in the request body (this agent’s ID), or use a “to” address like reply+AGENT_ID@yourdomain.com so we can derive the agent from the address.For domain/MX setup (e.g. receiving at your domain), see your provider’s docs or the project README.
Run history, notes, and usage for the agent you have loaded. Sign in and load an agent from the list above to see data.
Load an agent to see run stats.
Per-agent totals for the last 7 and 30 days: runs, reliability, speed, tokens, and estimated USD cost (model pricing approximation—not an invoice).
Est. cost uses token counts × default $/1M rates; your provider bill may differ.
Load a saved agent to generate your daily digest.
Run this agent to populate trend and ROI movement.
Bulk actions run immediately unless a guardrail countdown is active.
Outcome proof for this agent: time saved, autopilot confidence, approvals, and baseline guard status.
Run this agent to generate your weekly wins recap.
Queue drafts for approval-first publishing.
Queue health: waiting for drafts.
Queue SLA: no pending drafts.
SLA forecast: on track.
Today: 0 approved · 0 published.
Daily digest: no pending drafts.
Queue reliability (7d): n/a.
Queue trend (7d): -------
Recovery mode: off.
Recovery runbook: complete these steps to restore healthy queue flow.
Recovery ETA: estimating…
Recovery coach: target not available yet.
Auto-actions: off.
Dry-run preview: no stale items targeted right now.
Auto-action timeline: no attempts this session.
Breach snapshot: none captured this session.
Snapshot diff: waiting for breach baseline.
Recovery certificate: not generated yet.
Recovery leaderboard: no completed incidents yet.
Reminder: queued drafts are waiting too long.
Failure clusters: none.
Reward loop: concrete wins, risks, and next best action from this agent's last 7 days.
Add text or paste docs. The agent will use this in context when relevant. Save the agent first, then add chunks here.
TWITTERAPI_IO_KEY on the server)
No knowledge chunks yet. Add FAQs or docs above so the agent can answer from your content.
Curated skill packs you can attach to this agent. Packs add short, operator-grade instructions to the system prompt (stored on the blueprint as
runtime.installedSkills).
Draft-first posture for email side effects; explicit approval bias; leakage caution.
Lightweight weekly rhythm + platform-native hooks/CTAs; avoids engagement bait.
Owners/dates explicit; short follow-ups after meetings; fewer dropped commitments.
No packs installed yet.
Eight expert lenses run in parallel (no tools), then a synthesizer merges them into one prioritized digest. Add context from your business so advice stays grounded.
No digests yet. Run the council on a topic above.
When your agent needs approval (e.g. to send email or run a webhook), items appear here. Approve or dismiss each one.
No pending actions. Actions that require your approval will show up here.
Track who you contacted, status, and next follow-up. Add entries manually or paste Ivy’s output.
No outreach entries yet.
Unified contact view from email + outreach signals. Ask in plain English or queue a relationship nudge.
No contact profiles yet. Run inbox/admin workflows to build profile signals.
Capture meeting transcript/summary, extract action candidates, then approve to route to your tasks or waiting-on list.
No meetings captured yet.
Approve routes by owner: mine → Todoist (if connected), otherwise a calendar draft in Pending actions; theirs → Waiting on. Use Reject to dismiss. Approved items stay here for follow-through (mark done or snooze). The list aggregates actions across recent meetings when you refresh the dashboard.
No extracted action candidates yet. Capture a meeting and extract actions first.
Track commitments owned by others and nudge/check-in from one place.
No waiting-on items yet.
Outcome signals load when you refresh.
Connector health loads when you refresh the dashboard.
No failed runs in recent history.
Fathom: send webhook-id + timestamp headers for replay protection; same externalMeetingId updates the existing meeting instead of duplicating.
Webhook receipts: --
Operator view of the same run history feed: filter runs, inspect tool traces, and copy technical run details when support asks for them.
Includes only runs currently visible in this filtered list. What is JSON?
No runs match these filters.
No runs yet. Test your agent in the chat above; each run appears here with input, reply, and tools. Use Copy run summary to share or document.
Tip: Load an agent and click Send in the test chat to log your first run.
No notes yet.
Save durable preferences, constraints, and workflow facts your agent should remember.
No memory items yet.
No suggestions yet. Run your agent and click suggest.
Audit log = what happened. Facts & entities = info we inject into every run (e.g. your timezone, key people).
No entries yet.
Short facts we add to every run (e.g. timezone, company name). The agent can use these in its replies.
A list of people or companies the agent should know about. Use JSON like: [{"type":"person","name":"Jane","attrs":{"role":"CEO"}}]
Save version captures a named snapshot of this agent’s blueprint (Starter+). Restore rolls the agent back to that snapshot. Compare shows a JSON diff between two snapshots—use it before big prompt or tool changes. Pair with eval baselines for production safety.
No versions yet. Save a version or publish to create one.
Use a key to run this agent without logging in (e.g. X-API-Key header).
No keys yet.
Run a real homepage scan to check status, broken links, and script health.
Run Website QA in the background on a cadence and get email notifications.
No Website QA automation rules yet.
No Website QA runs yet.
Log in to connect Google Calendar.
Make inbox triage obvious: connect Gmail, load Robin, run one safe draft test, then approve sends from Pending actions.
Log in, then connect Gmail to start.
Make scheduling obvious: connect Google Calendar, load Jamie, run one draft planning test, then approve event actions safely.
Log in, then connect Google Calendar to start.
Connect Google and grant Gmail send permission to send from your Gmail account.
Sent emails are automatically spell- and grammar-checked before delivery.
Connect Outlook to send from your Microsoft 365 / Outlook account.
Sent emails are automatically spell- and grammar-checked before delivery.
Connect Calendly so your agent can share booking links. Direct add/change calendar events should use Google or Outlook tools.
After connecting, add Calendly or scheduling to your agent’s tools so it can offer booking links when users ask to meet. For direct date changes/invites, use Google or Outlook integrations.
Connect Zoom so your agent can create meeting links and include them in scheduling workflows.
When connected, add Zoom to your agent tools. The agent can create a Zoom meeting link, then draft calendar + email actions behind approval.
Connect Proofpoint, Mimecast, or Barracuda for security event triage workflows.
No security connectors configured yet.
No alerts loaded yet.
Connect a remote MCP server endpoint (Streamable HTTP). Then allow specific MCP tools per agent in Model & settings.
Calendly MCP: not connected.
Calendly MCP uses OAuth 2.1 + PKCE + Dynamic Client Registration (DCR). FreeAgentMaker connects you through the browser consent flow.
No MCP servers connected yet.
Connect Fathom webhook/API credentials for automatic meeting ingestion. Webhooks should include webhook-id and timestamp headers; duplicate externalMeetingId updates the same meeting.
Webhook URL: --
Connect Todoist to auto-dispatch approved “mine” meeting actions. If Todoist is missing or the API errors, we fall back to a calendar draft in Pending actions and show the reason in the status line.
Send event or meeting invites to a list of emails. Keep Dry Run on to preview safely.
CSV headers supported: email, firstName, lastName,
company.
Draft logos, signs, ads, brochures, labels, cards, posters, invites, and more (Gemini). Requires GEMINI_API_KEY. Approve before print or publish.
Upload images and videos for social posting. The publish tools use the public URLs.
No media yet. Upload something to start.
Paste an incoming email to instantly get priority, summary, draft reply, and next action.
-
-
-
-
Start free, then upgrade as your usage grows. Starter is $9/month, Pro is $19.95/month, and Premium is $39.95/month (public list prices; yearly may be offered at a discount). Pro includes higher production ceilings than Starter (agents, runs, voice) plus A/B compare in the test suite; Premium adds avatar builder features on top of Pro-level quotas. Compare the bullets in each card below. Usage limits match our Usage Policy and server enforcement (UTC). If you exceed included monthly usage on paid plans, we use soft caps: we pause and ask for explicit overage approval (once per UTC month) with an optional monthly budget cap.
Free
/ month
Most Popular
$9.00
/ month
Public list price for Starter unless we notify you otherwise.
$19.95
/ month
Public list price for Pro unless we notify you otherwise.
Custom
pricing
Seats, usage, SSO/BYOK/SLA, and annual terms are quoted in writing.
Team is a negotiated workspace tier (multi-seat, shared governance). It is not sold on the public checkout page—contact us for packaging. See Pricing for a short summary.
Founder teams use agents for support triage, lead qualification, outbound invites, meeting follow-through, website QA checks, and fast content repurposing. Start with one high-volume workflow, lock down guardrails, then expand.
Start with Support Triage. It reduces repetitive inbox work immediately and gives you faster first responses without hiring. Use the template, add escalation rules, then test with real customer messages before publishing.
Usage is measured by agent activity and run volume. Check your usage panel to see current totals, then set a budget and review thresholds before enabling higher-volume automations. If you hit limits, the app shows clear prompts so you can approve overage, reduce run scope, or wait for reset.
Your account controls your agent data and connected integrations. Keep sensitive workflows in draft-before-send mode, use least-privilege integrations, and review policy links in Privacy and Terms for current retention and provider handling details.
Yes. For email, CRM updates, publishing, and other high-stakes actions, start with approvals on by default. Approve every action at first, then gradually relax approval only for low-risk, reversible steps once quality is proven.
Ground agents in approved knowledge and constraints, not random public context. Add tone rules, require source-backed responses for factual claims, and escalate uncertain cases to a human. Run eval suites before turning on autopilot.
Use the Lead Qualification template and define ICP signals in constraints. Require output with score, reason, and next action so your follow-up stays human and focused. Validate against 10-20 real leads before making it part of your daily flow.
Yes. In Test Your Agent, open Send Invitations, paste your recipient list, and keep Dry Run on first. Review the preview, then send for real through your connected email provider once the copy and list look right.
Keep QA checks read-only, scope to approved URLs, and run simulation first. Add constraints that block destructive actions and require issue reports with URL, timestamp, and reproduction steps so fixes are fast.
Start with run logs and tool traces to identify the failing step. Most failures are auth, bad input shape, or external API limits. Fix one issue at a time in test mode, then rerun a small batch before restoring full volume.
Save a named version before major prompt, tool, or guardrail edits. Test the new version against your baseline scenarios; if quality drops, restore the previous version instantly. Version labels like "before FAQ update" make rollback fast.
Give the agent your tone rules, CTA style, and one to two example posts. Then generate post, email, and blog variants in one pass. Keep the highest-performing prompt recipe as your reusable founder playbook.
Social Media Manager Studio
Generate an approval-first weekly social plan: platform-native post drafts, engagement reply templates, hashtags, and a publishing checklist.
Visuals use Google Gemini image generation (Nano Banana). Sign in required. Generates up to 6 draft images from your post drafts.
Generate draft posts, then click Generate visuals (Nano Banana) to create images for each draft.