Question 1

What is the Smart Report Generator?

Accepted Answer

The Smart Report Generator is an advanced document processing tool that helps businesses quickly produce professional reports. Through AI technology, it automatically organizes data, applies templates, and generates reports that meet industry standards.

Question 2

What are the main features of the Intelligent Assistant System?

Accepted Answer

The Intelligent Assistant System provides real-time professional consultation services, including document review, content optimization suggestions, and best practice sharing. The system integrates expert knowledge from various fields, ensuring the accuracy and practicality of recommendations.

Question 3

How does the Report Generator improve work efficiency?

Accepted Answer

Through automated data processing and document generation workflows, it significantly reduces manual operation time, allowing teams to focus on more valuable analytical work, thereby improving overall work efficiency.

Question 4

What are the long-term benefits of using these tools for businesses?

Accepted Answer

Long-term use can enhance team productivity, ensure document quality consistency, reduce human error, and establish a corporate knowledge base, strengthening the organization's core competitiveness.

Question 5

How does the Report Generator ensure content accuracy?

Accepted Answer

The system employs multiple validation mechanisms, integrates the latest industry standards and regulatory requirements, and uses AI technology for content checking to ensure that generated reports are accurate and compliant.

Question 6

How does the Intelligent Assistant help improve report quality?

Accepted Answer

The Intelligent Assistant provides real-time content review and optimization suggestions, helping users identify potential issues and offering professional improvement solutions to ensure the overall quality of reports.

Question 7

Are these tools suitable for non-professionals?

Accepted Answer

Yes, the system design emphasizes user experience, providing an intuitive operating interface and clear usage guidelines, making it easy for users at all levels to get started.

Question 8

What types of reports can the Report Generator support?

Accepted Answer

The system supports various report types, including financial reports, business reports, project reports, etc., and can customize templates according to business needs to ensure different reporting requirements are met.

Question 9

How do these tools integrate with existing systems?

Accepted Answer

The system provides standardized API interfaces that can seamlessly integrate with existing corporate document management systems and workflows, enabling data sharing and process automation.

Question 10

Does the system regularly update content and features?

Accepted Answer

Yes, the system regularly updates features and knowledge bases to ensure compliance with the latest industry standards and best practices. Our expert team continuously optimizes the system to provide a better user experience.

Question 11

What is a workflow?

Accepted Answer

A workflow is a repeatable job you design once and run whenever you need it: the steps, the virtual employees who carry them out, the quality checks and the approval gates. You can generate one from a plain description of your scenario, start from a ready-made template, or edit the canvas directly.

Question 12

How do I tell a workflow what to ask for when it runs?

Accepted Answer

Open the workflow and use the input-parameters tab to declare what each run needs: a label, a type, whether it is required and an optional default. Anyone who runs the workflow then gets a form with exactly those fields, and a run that is missing a required value is rejected before it starts spending credits.

Question 13

Can a workflow read files that I upload when the run starts?

Accepted Answer

Yes. Declare a file parameter and the run dialog shows an upload field; single-file and multi-file inputs are both supported, and a knowledge base can be selected the same way. Uploads go to your own storage and count towards your plan's storage quota, so an oversized file is refused with the limit stated instead of failing mid-run.

Question 14

Where do a workflow's results go, and can I download them?

Accepted Answer

Every run keeps a complete record: each step's output, the files it produced, and the tokens and credits it consumed. Files produced by a run appear in the run's results with a download link, and a report created by a workflow also shows up in your normal report list.

Question 15

Can I start a workflow from my own system?

Accepted Answer

Yes. Create an external endpoint for the workflow, then call it with the endpoint secret to start a run, poll its status, and supply an answer while the run is waiting for input. The endpoint page lists the exact parameters that workflow expects and gives you a ready-to-copy request example.

Question 16

Can workflows run on a schedule?

Accepted Answer

Yes. Add a schedule trigger and choose the frequency and time zone. Fixed values for the workflow's input parameters are stored on the trigger, so a scheduled run never stops to wait for a person. If a scheduled run keeps failing for the same reason, the schedule is paused and the owner is notified instead of failing silently.

Question 17

Where do I see or change my plan?

Accepted Answer

Your plan, renewal date and included quotas are in account settings. An upgrade takes effect immediately; a downgrade or cancellation applies from the next billing cycle. Features that are no longer included become read-only rather than deleting anything you already created.

Question 18

How are credits consumed, and what happens when they run out?

Accepted Answer

Credits are consumed by model usage — chats, report generation and workflow runs — and the amount depends on the model and the size of the request. Usage is itemized in your account, and when the balance reaches zero a new run is refused up front with a clear message instead of stopping halfway through.

Question 19

How do I share what I build with my team?

Accepted Answer

Members of the same organization can be given access to agents, teams, workflows and knowledge bases — individually, by department, or organization-wide. The owner stays in control: shared items can be opened and run by others, while editing and deletion remain with the owner unless write access is granted explicitly.

Question 20

How do I get an endpoint secret, and how should I protect it?

Accepted Answer

Create an external endpoint for the resource you want to expose; the secret is shown once at creation and can be rotated at any time. Keep it on your server side only — anyone holding it can start runs billed to your account — and rotate it immediately if it may have been exposed.

Question 21

Are there rate limits on the API?

Accepted Answer

Yes. External endpoints are rate limited per endpoint and per organization; requests over the limit are refused with a 429 and a retry hint rather than being dropped silently. Long jobs are asynchronous: you start a run, receive its id, and then poll for its status.

Question 22

How do I verify that a callback really came from the platform?

Accepted Answer

Callbacks are signed. Each request carries a timestamp and a signature computed over the raw request body with your endpoint secret; recompute it on your side and compare before trusting the payload, and reject requests whose timestamp is too old so a captured callback cannot be replayed.

Question 23

How do I connect Claude Code to Akashic?

Accepted Answer

Akashic is a remote MCP server, so Claude Code can call your virtual employees, workflows, reports and knowledge bases directly. Create an API key with the MCP tools permission under `/agent-workforce/endpoints`, then open `/agent-workforce/connect` and copy the Claude Code snippet: it runs claude mcp add --transport http akashic YOUR_SERVER_URL --header "Authorization: Bearer YOUR_API_KEY" with the real URL and key already filled in. To share the connection with a project instead, use the .mcp.json snippet from the same page. Run claude mcp list to confirm the server shows as connected, then ask Claude to list your virtual employees.

Question 24

How do I connect Codex CLI to Akashic?

Accepted Answer

Codex reads MCP servers from ~/.codex/config.toml. Add a [mcp_servers.akashic] section with url set to your Akashic MCP endpoint and bearer_token_env_var = "AKASHIC_API_KEY", then export the AKASHIC_API_KEY environment variable with a key that has the MCP tools permission; the key never has to be written into the file. Alternatively run codex mcp add akashic --url YOUR_SERVER_URL --bearer-token-env-var AKASHIC_API_KEY. The Codex snippet under `/agent-workforce/connect` already contains your endpoint URL, and keys are created under `/agent-workforce/endpoints`.

Question 25

Why does a tool reply "approval required", and where do I approve it?

Accepted Answer

Tools that change something — creating a report, starting or cancelling a workflow, submitting an answer to a paused run, drafting an email — are parked until a person approves them. Your client receives an approval request id and an expiry time instead of a result. Open `/background-tasks`, find the call in the awaiting-approval list and choose approve or reject; the notification bell also reminds you of pending calls. A request must be approved within 24 hours, after which it expires and the client has to send it again. Once approved, the tool runs and the client can fetch the result with get_approval_status.

Question 26

Why does my MCP client only see chat tools?

Accepted Answer

The tool list is computed from the API key. If the key was created without the Platform tools permission, only the generic tools are exposed: list agents, list teams, and chat with an agent or a team. Edit the key under `/agent-workforce/endpoints`, enable Platform tools and tick the tools it may call; the developer preset does this for you. If tools stay hidden, your organization admin may have turned off external tool access for the organization, or the platform may not expose that tool yet. Reconnect or restart the client so it fetches the updated tool list.

Question 27

How do I let a key run write tools without waiting for approval?

Accepted Answer

Auto-approval is decided by three layers, and every layer has to agree. First, the platform sets a floor: some tools always require a person. Second, an organization admin allows specific tools in the organization policy under `/org-admin`; a personal account without an organization skips this layer. Third, the key itself carries a no-approval list, which you set when creating or editing the key under `/agent-workforce/endpoints`; only tools the higher layers allow can be selected. The connect page shows the effective outcome per tool, so you can see exactly why a call still asks for approval.

Question 28

Which workflow design pattern should I pick?

Accepted Answer

Start from what the job needs. One-shot data tidying or summarisation: Augmented LLM. Multi-step reasoning with tools: Reflection loop (ReAct) or Tool-use agent. Requests of several kinds: Router. Long, decomposable work: Planner-executor. Writing that must pass a quality bar: Evaluator-optimizer. Payments, approvals or compliance: Verifier-gated. Cross-team analysis over shared knowledge: Multi-agent with memory. Critical answers that must be right: Self-consistency (ensemble). Personalised or long-running support: Memory-augmented. If you are unsure, do nothing: the pattern selector lives under "Advanced settings" in the composer and stays on Automatic until you open it and pick one.

Question 29

What is the difference between generating a workflow with AI and creating a skeleton?

Accepted Answer

Generating with AI reads your scenario and builds a workflow in the chosen pattern; it costs credits like any AI generation and the AI fills in prompts, roles and inputs for you. A skeleton is created instantly for free from a fixed blueprint of the pattern with generic prompts; it counts toward your workflow limit and you complete it on the canvas, where the AI assistant can help. Open "Advanced settings", pick a pattern and leave the scenario empty to get a skeleton.

Question 30

Why does the canvas say my workflow deviates from its design pattern?

Accepted Answer

A workflow created from a pattern carries a pattern tag. The canvas checks the structure against that pattern (for example an evaluator-optimizer needs a loop that ends with a verify step) and warns when something is missing or when a knowledge base or tools still need to be bound. Warnings never block saving or running. If you deliberately changed the architecture, remove the pattern tag from the badge in the header.

Frequently Asked Questions