Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
102 changes: 51 additions & 51 deletions .github/twins.lock.json

Large diffs are not rendered by default.

10 changes: 6 additions & 4 deletions account-management/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -58,21 +58,23 @@ Settings > Members > click the **✕** icon next to the member.
| Plan | Price | AI Credits/mo |
|------|-------|--------------|
| **Free** | $0 | 0 (earn credits with referrals and invites) |
| **Starter** | $6/mo, billed annually | 10,000 |
| **Starter** | $6/mo, billed annually (yearly only) | 10,000 |
| **Pro** | $10/mo billed annually, or $20 monthly | 50,000 |
| **Business** | $25/mo billed annually, or $50 monthly | 150,000 |
| **Max** | $100/mo billed annually, or $200 monthly | 400,000 |
| **Enterprise** | Custom (contact sales) | Custom |
| **Enterprise** | $250/mo billed annually, or $500 monthly (contact sales) | 1,100,000 |

Key features of each plan:

- **Free**: 2 members, 3 Taskade Genesis apps, 1 AI agent, 10 automation runs a month.
- **Free**: 2 members, 3 Taskade Genesis apps, 1 AI agent, 10 automation runs a billing period.
- **Starter**: 3 members, unlimited apps, premium integrations, hosted MCP access.
- **Pro**: 10 members, custom domains (up to 10), webhooks, your own AI key (BYOK), app bundle export.
- **Business**: Unlimited members, unlimited custom domains, SSO/SAML, app analytics, custom branding.
- **Max**: Everything in Business, with higher AI limits.
- **Max**: Everything in Business, with higher AI limits. Business allows 80,000 AI requests a day, and Max allows 320,000.
- **Enterprise**: Everything in Max, plus SCIM provisioning and the highest AI limits.

Automation runs a billing period are 10 on Free, 350 on Starter, 1,000 on Pro, and 10,000 on Business and up. The whole workspace shares the Monthly Credits and the automation runs. The table shows the base level of each plan. A higher **credits / month** level on the upgrade page raises both.

Full details: [taskade.com/pricing](https://www.taskade.com/pricing)

### Manage Your Subscription
Expand Down
131 changes: 70 additions & 61 deletions account-management/credits-and-billing.md

Large diffs are not rendered by default.

65 changes: 40 additions & 25 deletions apis-living-system-development/agent-embeds.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,27 +5,27 @@ description: >-

# Embed a Public AI Agent

Add a public Taskade AI agent to your website with an iframe or a chat widget. You publish the agent, copy the code from **Publish** → **Embed**, and paste the code into your page.
You can put a public Taskade AI agent on your website as an iframe or as a chat widget. Publish the agent, copy the code from **Publish** → **Embed**, and paste the code into your page.

## Before you start

You need an AI agent and edit access to your website's HTML.
You must have these items:

- **An AI agent in your workspace.** Publishing and embedding are on every plan.
- **Access to your site's HTML.** You paste one `<iframe>` or two `<script>` tags.
- **Optional, Business and above:** **Hide Branding**, a custom background, and a meta image for the public agent.
- **An AI agent in your workspace.** Every plan can publish and embed an agent.
- **Edit access to the HTML of your website.** You paste one `<iframe>` or two `<script>` tags.
- **Optional, Business and up:** **Hide Branding**, a custom background, and a meta image for the public agent.

## Publish the agent and copy the code

Turn on public access, then copy the code that Taskade generates for you.
Turn on public access. Then copy the code that Taskade makes for you.

1. Open the agent, then open the **Publish** tab. In the agent list, the **Share** menu item opens the same screen.
1. Open the agent, then open the **Publish** tab. The **Share** menu item in the agent list opens the same screen.
2. Turn on the **Publish Agent** switch.
3. Open the **Embed** sub-tab.
4. Pick **Add to your website** for an iframe, or **Add a chat widget** for a widget.
5. Copy the generated code.

The generated code holds the current public agent ID and only the options that the selected mode supports. The **Customize** sub-tab sets the look. **Tool Permissions** shows only when the agent has tools.
The code holds the current public agent ID. It holds only the options that the selected mode supports. The **Customize** sub-tab sets the look. **Tool Permissions** shows only when the agent has tools.

```mermaid
graph TD
Expand All @@ -38,7 +38,7 @@ graph TD

## Choose an embed type

Pick the smallest surface that fits the page.
Pick the smallest embed that fits the page.

| Embed | What visitors see | Best fit |
| --- | --- | --- |
Expand All @@ -49,19 +49,19 @@ Pick the smallest surface that fits the page.
| `searchBar` | An Ask AI pill mounted inside your layout | Headers and documentation |
| `inline` | An open chat panel mounted inside your layout | Dedicated assistant sections |

`bubble` is the widget default. Taskade leaves it out of the generated config.
`bubble` is the widget default, so Taskade does not put it in the generated code. An agent with saved widget settings is different. Read [Save the widget settings as the default](#save-the-widget-settings-as-the-default).

`inline` must mount into a page element. `searchBar` can mount into one, for a pill in a site header. The generated snippet includes this host element:
An `inline` widget must mount into a page element. A `searchBar` widget can also mount into one, for a pill in a site header. The generated snippet includes this host element:

```html
<div id="taskade-agent"></div>
```

Move this element to the place where the pill or panel must appear. Keep its ID aligned with the `target` value. A `searchBar` without a `target` is a pill fixed to the corner that `popup.position` sets.
Move this element to the place where the pill or panel must show. Keep its ID the same as the `target` value. A `searchBar` with no `target` is a pill in the corner that `popup.position` sets.

## Add an iframe

Use the iframe when the agent must stay open inside a known area.
Use the iframe if the agent must stay open inside a fixed area of the page.

```html
<iframe
Expand Down Expand Up @@ -92,9 +92,9 @@ Place the generated script near the end of the page body.
</script>
```

Keep the double `TaskadeEmbed.TaskadeEmbed.init` name. The CDN bundle exposes the current initializer at that path.
Keep the double name `TaskadeEmbed.TaskadeEmbed.init`. The CDN bundle gives the current initializer at that path.

The snippet stores the widget handle on `window.taskadeAgent`. Your page code can call these methods:
The snippet keeps the widget handle on `window.taskadeAgent`. The code of your page can call these methods:

| Method | Effect |
| --- | --- |
Expand All @@ -117,6 +117,14 @@ For an inline panel, add the host element. Then set the mode and the target:
</script>
```

### Save the widget settings as the default

**Save as default** on the **Embed** sub-tab stores the widget settings on the agent. You can save a default for the `bubble`, `modal`, and `sidebar` modes. A `searchBar` or `inline` widget sits inside your page, so its settings stay in the snippet.

If the settings on the screen match the saved default, Taskade makes a short snippet that holds only `publicAgentId`. The widget loads the saved settings when it starts. A `trigger` or `interface` option stays in the snippet, because Taskade does not save those two. A later save updates every site that uses the short snippet.

If a snippet sets one of `mode`, `theme`, `popup`, `launcher`, `behavior`, or `visibility`, the widget ignores all saved settings. To show the plain widget on one site, set `"mode": "bubble"` in the snippet of that site.

## Configure the widget

Pass only the options that the selected mode uses.
Expand Down Expand Up @@ -144,42 +152,49 @@ Pass only the options that the selected mode uses.

Some options apply only to some modes:

- The `theme` radius, density, and font style the widget around the chat, not the chat transcript.
- The `theme` radius, density, and font change the widget around the chat. They do not change the chat transcript.
- `launcher.size` does not apply to `searchBar` or `inline`.
- `popup.position` applies to `searchBar` only when it has no `target`.
- `inline` is always open. It uses no label, proactive message, auto-open delay, custom trigger, or hotkey.
- An `inline` widget is always open. It uses no label, proactive message, auto-open delay, custom trigger, or hotkey.

## Keep public access safe

Treat the embed as public access to the agent.
The embed gives the public access to the agent. Do these steps before and after you publish:

- Review the agent's instructions and knowledge before you publish it.
- Keep secrets and private records out of public knowledge sources.
- Turn off tools that a public visitor must not use.
- If the public agent does not need the web, keep **Search the web** and **Read a web page** off for visitors.
- Test the public page in a signed-out browser.
- Turn off the **Publish Agent** switch if the link reaches the wrong audience.
- Keep your API token out of the page. The embed code needs only the public agent ID. Your own server code uses a token, as [Authentication](developers/authentication.md) shows.
- If the wrong people get the link, turn off the **Publish Agent** switch.
- Keep your API token out of the page. The embed code needs only the public agent ID. Only your server code uses a token. Read [Authentication](developers/authentication.md).

{% hint style="info" %}
**Note:** A public agent keeps no memory between visitor chats. Read what visitors asked in the agent's **Messages** tab.
{% endhint %}

A public chat can search the web or read a page only when the matching tool is on for visitors. A visitor does not see tool results or the reasoning of the agent. Taskade also hides the input of most tool calls from the visitor. The visitor sees the reply and the web sources of a search. A signed-in member of the workspace of the agent still sees the tool details.

## Review questions the agent did not answer

The Agent Inbox collects the public questions that your agent did not answer well, so you can teach it the answer. It covers chats on the public agent page and in the embed. A private chat in your workspace never goes to the inbox.
The Agent Inbox collects the public questions that your agent did not answer well. You can then teach the agent the answer. The inbox covers chats on the public agent page and in the embed. A private chat in your workspace never goes to the inbox.

Two things add a question to the inbox:

- After the agent replies in a public chat, Taskade checks the chat for a question that the agent did not answer.
- A visitor clicks the thumbs-down button (**Bad response**) on a reply. The question behind that reply goes to the inbox, with the reason that the visitor typed. The reason box tells the visitor that the agent owner can see it.
- After each agent reply in a public chat, Taskade looks for a question that the agent did not answer.
- A visitor clicks the thumbs-down button (**Bad response**) on a reply. The question of that reply goes to the inbox with the reason that the visitor typed. The reason box tells the visitor that the agent owner can read it.

To review the questions:

1. Open the agent, then open the **Knowledge** tab.
2. Open the **Inbox** sub-tab. It shows after Taskade adds the first question.
3. For each question, click **Answer & save**, write the answer, then click **Save as knowledge**. To skip a question, click **Dismiss**.
3. For each question, click **Answer & save**.
4. Write the answer.
5. Click **Save as knowledge**.

To skip a question, click **Dismiss**.

The saved answer becomes agent knowledge, so the agent can use it with the next visitor. Taskade also keeps the questions in a workspace project named `Agent Inbox - <agent name>`, which every member who can view agents can read.
The saved answer becomes agent knowledge, so the agent can use it with the next visitor. Taskade also keeps the questions in a workspace project named `Agent Inbox - <agent name>`. Every member who can view agents can read this project.

| Limit | Value |
| --- | --- |
Expand Down
45 changes: 40 additions & 5 deletions apis-living-system-development/autonomous-agents.md
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,7 @@ A schedule does not start at the exact minute that you set. Taskade spreads the
| Every hour, or less often | 30 minutes |
| More often than every hour | Half the interval, at most 10 minutes |

So a run set for **Every day** at 09:00 starts between 09:00 and 09:30. Do not build an automation that needs the run to start at an exact minute.
Thus a run set for **Every day** at 09:00 starts between 09:00 and 09:30. Do not make an automation that must start at an exact minute.

## Pattern: Event-Triggered Agent

Expand Down Expand Up @@ -135,7 +135,7 @@ The step does not support Agent App tools. If a tool needs manual approval, the

A team member can have a role, for example Researcher or Writer. When you edit a team, click **Add role** next to a member. A role is one line of at most 60 characters.

The team manager reads the roles when it picks the member for each task or step. In **Orchestrate** mode, Taskade tells the manager to give a member with the role `Verifier` (in any letter case) review steps only, never a step that creates, changes, or sends data. This is an instruction to the manager, not a tool limit: the member keeps its tools. A team with no roles works as before.
The team manager reads the roles when it picks the member for each task or step. In **Orchestrate** mode, Taskade tells the manager to give only review steps to a member with the role `Verifier` (in any letter case). The manager never gives this member a step that creates, changes, or sends data. This is an instruction to the manager, not a tool limit. The member keeps its tools. A team with no roles works as before.

You can also ask Taskade EVE to rename a team, change its instruction, add or remove a member, or set a role.

Expand Down Expand Up @@ -209,11 +209,31 @@ The response is the same shape as any `promptAgent` call:
}
```

Cross-agent invocation moves the orchestration into Taskade. Your app sends one request, and the automation does the handoff.
With cross-agent invocation, Taskade does the orchestration. Your app sends one request, and the automation gives the work to the next agent.

## Pattern: Human Approval

A **Wait for approval** step pauses a run until a person approves or rejects the request. Use it after an agent step and before an action that you cannot undo, such as a payment or a message to a customer.

| Setting | What it does |
| --- | --- |
| **Who approves** | The workspace members that you pick. The first answer counts |
| **What they see** | A short line that says what to approve |
| **Details** | Values from earlier steps that help the approver decide |
| **Wait for** | 1 day, 3 days, or 7 days. The default is 3 days |
| **Then** | **Approve** or **Reject** when the wait ends with no answer. The default is **Reject** |

The approver gets an email and can decide in Taskade. An approval continues the run. A rejection, or a timeout set to **Reject**, stops the run.

Later steps can read the decision fields `approved`, `decision`, `decidedBy`, `decidedAt`, and `comment`. The `decision` field is `approved`, `rejected`, or `timed_out`. While the run waits, its status is `waiting`. The `get_flow_runs` tool of the [hosted Taskade MCP](genesis-app-mcp.md) reports the same status.

A connector can also send the request in its own channel. Gmail, Outlook, Slack, Microsoft Teams, Telegram, Discord, Twilio (SMS), and WhatsApp have request-approval actions. These connectors need a paid plan. A request-approval action waits 3 days, then rejects if no one decides.

An agent does not get a request-approval action as a tool. If an agent calls one, Taskade refuses the call.

## Monitoring and Observability

Watch each run, and let Taskade stop an automation that keeps failing.
Watch each run. Taskade stops an automation that fails again and again.

### Run history

Expand All @@ -223,7 +243,7 @@ The automation's **Runs** tab shows the **Run History**. Filter it with **All**,

AI steps use credits. These screens show where the credits go:

- **Settings** → **Credits**: the **Detailed Usage Breakdown** for the account.
- **Settings** → **Credits**: **Usage this month** for the account, with the **Where your credits went** list for the person who pays for a paid plan.
- **App Settings** → **Insights** → **AI Usage**: the credits, tokens, and builds for one app.

### Continue after a failed step
Expand All @@ -237,6 +257,21 @@ An action step has a **Continue if this step fails** setting. Turn it on for a s

With either choice, Taskade does not mark the run failed, and the failed step still shows in **Runs**. This also applies to a step inside a **Choose a path** branch.

### Retry a step when a service is down

An action step has a **Retry if this step fails** setting. Use it for a step that calls a service that can be busy or down for a short time.

| Setting | Values |
| --- | --- |
| **Wait** | 1 to 5 minutes between tries. The default is 5 minutes |
| **then try again, up to** | 1 to 5 more tries. The default is 3 |

Taskade retries only a failure that a later try can fix. A step with bad input or a broken connection fails at once, with no retry. **Continue if this step fails** applies after the last try. Agent steps, such as **Ask an agent** and **Ask an agent team**, do not have this setting.

{% hint style="warning" %}
**Warning:** If a step sends a message, turn on retries only if a second copy is permitted. A retry can send the message twice. Retries can also delay later runs of the automation.
{% endhint %}

### Automatic pause after failures

Taskade turns off an automation that fails again and again. This stops a broken connection from using more runs.
Expand Down
Loading
Loading