Feature Guide
Personal Autonomous Agent Platform
This guide covers what Prosponsive can do and why each capability matters. Every feature described here exists because existing AI tools force you into a pattern that wastes your time, leaks your data, or locks you into a vendor. Prosponsive was built to fix that. If you haven't used Prosponsive yet, start with the Install Guide and Basics.
Prosponsive executes tools asynchronously. When an agent calls a tool, the work happens in the background. You do not sit and watch a chat window spin.
When you give an agent a task that involves tools, the agent submits tool calls and they execute independently. You can close the conversation, switch to another channel, or leave your desk entirely. When the tools finish, results are waiting for you. The agent has already incorporated them and moved on to the next step, or is ready for your input.
Most AI tools work synchronously: you type, the AI responds, you watch it think, you wait for the tool to run, you read the result, you respond. Your attention is captured for the entire duration of every tool call. If a tool takes 30 seconds, you lose 30 seconds of focus. If a workflow chains five tools, you lose minutes sitting idle in a chat window.
Prosponsive eliminates that pattern. You send a task and walk away. The difference between synchronous and asynchronous tool use is the difference between standing at a fax machine waiting for each page and dropping a stack in the outbox. One demands your attention. The other respects it.
Async tool use changes what you can ask an agent to do. When tools run in the background, it becomes practical to assign tasks that take minutes to complete — pulling data from multiple systems, generating reports, processing documents. These are tasks you would never assign through a synchronous chat interface because you would have to sit and watch the entire time.
Prosponsive channels are not one-on-one conversations with a single AI. They are group chats. Multiple agents share a channel, and when you post a message, any agent assigned to that channel can respond — independently, in parallel, and asynchronously — each using its own tools.
When you send a message to a channel with multiple agents, each agent processes it on its own terms. A research agent starts pulling data. A writing agent starts drafting. A monitoring agent checks the systems you asked about. None of them wait for the others. Each agent's tool calls run independently — and because tools execute asynchronously, each agent's work continues in the background while the others do the same.
You post once. Parallel work begins across all agents simultaneously. Results arrive as they complete.
Most AI products are one-on-one: you message one assistant, it responds, you respond back. The conversation alternates in strict turns. If you want multiple agents involved, you route between them manually — copy context from one conversation to another, or build an orchestrator that handles routing for you.
In Prosponsive, the channel is the coordination layer. All agents in a channel share the same conversation context automatically. No routing code. No context copying. You post a message and every agent in the channel sees it, responds on its own schedule, and posts results back — all asynchronously, all in parallel.
A single agent with all your tools can handle sequential tasks. But when multiple agents with specialized tool sets work in parallel, you get specialization, isolation, and concurrency at the same time.
Consider three agents in a channel: one with read access to your data sources, one with write access to your communication tools, and one that generates and formats documents. Each has only the tools it needs. The research agent pulls information. The communications agent drafts outreach. The document agent formats the output. All three can run simultaneously on a single message. One general agent with all three tool sets would run those tasks sequentially, carry a larger attack surface, and produce less focused results — because the model has to sort through everything to find what matters at each step.
The group chat model is what makes this practical without orchestration code. Set up the channel once, share the right agents to it, and post. The agents do the rest.
Prosponsive organizes work into channels and threads, not flat conversations. This structure affects token cost, prompt quality, how agents collaborate, and how you stay organized as work scales.
A channel is a persistent workspace where one or more agents operate. Agents and their tools are scoped to channels — an agent assigned to a channel has access to specific tools relevant to that channel's purpose. An agent not assigned to a channel cannot see it, cannot participate in it, and does not consume tokens for it.
In most AI products, every conversation has access to every tool and every capability. The system prompt grows with every integration you add. Token costs increase linearly, and the AI model gets a bloated context full of irrelevant instructions. Channels solve this by scoping agents and tools to where they belong.
Within a channel, individual tasks and topics live in threads. Threads keep related messages together without polluting the channel's main flow. When an agent processes a scheduled prompt, the result is a thread. When a tool call produces a detailed result, it lives in a thread. When a conversation branches into a subtopic, it becomes a thread.
Threads are what make channels usable at scale. Without them, a busy channel would be a wall of interleaved messages from different tasks and agents. With threads, each task has its own contained context. You can collapse threads you've reviewed, expand the ones that need attention, and keep a clear view of what's active. Agents can also move threads between channels — when a conversation surfaces a topic that belongs elsewhere, the thread moves to the right place without losing its history.
Because agents and tools are scoped to channels, each conversation carries only the context it needs. An agent managing your email does not receive instructions about your database queries. An agent that handles calendar scheduling does not see your code review tools. This keeps prompts lean, reduces token usage, and produces better results because the model is not sorting through irrelevant context to find what matters.
When you create an agent, you define which channels it can participate in. This means you control exactly which agents can see and respond in each workspace. You can share a channel across multiple agents for collaboration, or restrict it to a single specialist agent for focused work.
Prosponsive is built for tasks, not talk. When agents use tools, they generate actions that need your review. In a typical AI chat product, you see these one at a time in a linear conversation. Prosponsive gives you filtering, grouping, and bulk approval instead.
When multiple tool actions are pending, Prosponsive lets you filter them by agent, by tool, by channel, or by status. You can group related actions together — all the file operations in one group, all the API calls in another. This gives you a structured view of what your agents want to do, rather than a scroll of interleaved chat messages.
Once you have filtered and grouped actions, you can approve them in bulk. Review a set of related tool calls, confirm they look correct, and approve them all at once. Sequential chat interfaces present one action at a time embedded in a conversation thread, making bulk review impractical.
When you have multiple agents working across multiple channels with multiple tools, the volume of pending actions can grow quickly. Without filtering and bulk approval, you become a bottleneck — clicking "approve" one action at a time, scrolling through conversations to find what needs attention. Prosponsive's task-oriented design means you can review and approve a batch of actions in seconds instead of minutes.
Prosponsive supports condition-based auto-approvals. You define rules, and matching tool calls execute without interrupting you.
An auto-approval rule specifies conditions under which a tool call should be approved automatically. You can set conditions based on:
When a tool call matches an auto-approval rule, it executes immediately. When it does not match, it queues for your manual review like normal.
Auto-approvals are what turn Prosponsive from an interactive tool into an always-on system. Without them, every tool call blocks until you manually approve it. With them, trusted operations run continuously. An agent checking your inbox every 15 minutes via a scheduled prompt does not need you to click "approve" every time it calls the email-reading tool. You set the auto-approval rule once, and the agent operates independently from that point forward.
Auto-approvals are not all-or-nothing. You can auto-approve low-risk read operations while requiring manual approval for actions that modify data. You can auto-approve tools in a monitoring channel while keeping approval requirements in a channel that manages production systems. The rule system is granular enough to match your actual trust boundaries.
Agents in Prosponsive don't just answer questions — they manage your experience. They have built-in capabilities to organize, prioritize, and clean up after themselves so you spend less time managing and more time reviewing outcomes.
Agents can move threads to different channels. When a conversation spawns a topic that belongs somewhere else — a project discussion that surfaces a bug, a brainstorming session that produces an action item — the agent moves it to the right channel. Information ends up in one place regardless of where it originated, organized by context rather than by when it happened to come up.
When an agent finishes a task, it can archive the channel or messages it no longer needs. Completed work doesn't clutter your view or consume tokens as context in future messages. Agents clean up after themselves — they know when work is done and act on it.
Prosponsive gives you clear visibility into your agents' efficiency and effectiveness. You can see how your agents are performing — which ones are getting work done, which ones are consuming tokens without producing results, and where you have opportunities to save time and cost. This is not usage reporting for billing purposes. It is operational insight that helps you tune your agents, adjust your auto-approval rules, and identify where automation is working and where it isn't.
The combination of agent autonomy, auto-approvals, and performance visibility creates a feedback loop: agents work independently, you review their results in bulk, performance data shows you where to expand automation, and your agents become more effective over time with less input from you.
Prosponsive keeps you informed without demanding your attention. Notifications surface what matters, agent definitions control how agents behave, and scheduled prompts keep work moving — together they form the control layer that makes always-on operation practical.
You decide when, why, and what your agents notify you about. Prosponsive has no system notifications, no promotional alerts, no update banners, no engagement nudges. The only notifications that exist are the ones you and your agents create. Every notification is a signal you asked for — a tool action requires your approval, a scheduled task found something worth reporting, or an agent hit a condition you told it to flag. The result is a 100% signal-to-noise ratio, because you define what counts as signal.
Every agent in Prosponsive is defined by a markdown file that specifies its identity, instructions, tools, and behavior. Agent definitions are the control surface for how your agents operate — you define what each agent knows, what tools it can use, which channels it participates in, and how it should behave. This is not prompt engineering buried in a settings panel. It is a clear, editable document that fully describes your agent's capabilities and boundaries.
Agent definitions combine with channels, auto-approvals, and scheduled prompts to create a complete automation system. Define an agent's behavior in its definition, assign it to the right channels, set auto-approval rules for its tools, schedule recurring prompts to trigger it, and configure notifications for when it needs you. Each piece is simple on its own — together they give you fine-grained control over autonomous operation.
Consider how these pieces work together: A scheduled prompt fires every morning, sending a message to your reporting agent. The agent uses its tools — auto-approved because you trust them — to gather data from your systems. It posts the results to the reporting channel, archives yesterday's report, and notifies you only if something unusual needs your attention. You defined all of this through an agent definition, a few auto-approval rules, one scheduled prompt, and a notification preference. No code. No orchestration. Just configuration.
Custom tools are not MCP servers. They are fundamentally different in scope and design. Understanding this difference is key to understanding why Prosponsive exists.
An MCP (Model Context Protocol) server gives an AI model access to a single system — read from a database, post to Slack, create a GitHub issue. Each MCP is a thin wrapper around one API. To accomplish anything meaningful, you have to orchestrate multiple MCP calls yourself, or hope the AI model figures out the right sequence.
A Custom tool is an n8n workflow. A single tool can connect to multiple systems, apply business logic, transform data between steps, handle errors, and produce a finished outcome. One custom tool does what would require multiple MCP calls plus manual coordination to achieve.
Consider a tool that processes a customer support ticket. With MCPs, the AI would need separate calls to: read the ticket from your helpdesk, look up the customer in your CRM, check their subscription status in your billing system, draft a response, and post it back. That is five separate tool calls across five systems, with the AI model responsible for sequencing and data passing.
In Prosponsive, this is one tool. The n8n workflow handles the entire sequence: it reads the ticket, looks up the customer, checks billing, and returns a structured result to the agent. The agent makes one tool call and gets a complete picture.
Every custom tool is built in n8n, a visual workflow automation platform that runs locally alongside Prosponsive. You build tools by connecting nodes in a drag-and-drop editor. n8n has pre-built nodes for over a thousand services — Slack, Gmail, Google Sheets, Notion, GitHub, Jira, databases, HTTP APIs, and more. You can inspect, modify, and debug every tool. Nothing is a black box.
When an agent calls a tool, it never sees your credentials. Your API keys, passwords, and tokens are stored in n8n's encrypted credential manager. The agent sends a request to the n8n workflow, n8n executes using its stored credentials, and the agent receives only the result. The AI model never has access to your secrets — not in the prompt, not in the tool call, not anywhere in the conversation.
This is a meaningful security boundary. You can give agents access to powerful tools that interact with production systems, and your credentials remain completely isolated from the AI provider.
For a hands-on guide to building tools, see the Developer Guide.
Prosponsive supports 7+ AI providers, and you can switch between them at any time. But provider independence is not just about avoiding lock-in — it is about aligning incentives.
| Provider | Type | Key Characteristic |
|---|---|---|
| Ollama | Local | Fully offline, runs on your hardware |
| Anthropic | Cloud API | Claude models, strong reasoning |
| OpenAI | Cloud API | GPT models, broad ecosystem |
| Cloud API | Gemini models, multimodal | |
| Groq | Cloud API | Fast inference for open models |
| Mistral | Cloud API | European-hosted, efficient models |
| AWS Bedrock | Enterprise Cloud | Access multiple model families via AWS |
Changing your AI provider is a single setting change. Your conversations, agents, tools, and workflows are completely independent of which provider you use. Switch from OpenAI to Anthropic to Ollama and back — nothing else changes. You use your own API keys with each provider. Your AI requests go directly to the provider you choose.
When your AI provider goes down or you hit a rate limit, your work stops. If you depend on a single provider, an outage means your agents can't function, your scheduled prompts don't run, and your tools sit idle. Prosponsive solves this with automatic failover — you define a priority list of providers and models, and if the primary fails, Prosponsive switches to the next one without interruption. Your agents keep working, your tasks keep running, and you may not even notice the outage happened.
Most AI products are built by AI providers or depend on a single provider's revenue. Their incentive is to maximize your token consumption. Prosponsive's incentive is the opposite.
Every design choice in Prosponsive works to keep your token usage low. Channels scope agents and tools so prompts carry only relevant context. Agents receive only the instructions and tool definitions they need for the current channel, not a bloated system prompt containing everything. Group agent collaboration means specialist agents with small, focused prompts rather than one general agent with a massive context window.
The result: you get better outputs at lower cost because the AI model processes focused, relevant context instead of sorting through noise. Prosponsive serves you, not your token bill.
Prosponsive keeps your work private during execution. Your conversations, your data, and your tool activity all stay on your machine — not on someone else's server.
Prosponsive was built with enterprise security scrutiny by an engineering leader whose career was spent building software for some of the most security-conscious and heavily regulated organizations in the world. That background shaped every architectural decision — local execution, encrypted connections, credential isolation, and data sovereignty are not afterthoughts bolted on for compliance. They are foundational to how Prosponsive works.
When Prosponsive runs your tools, processes your conversations, and stores your data, everything happens on your computer. The database, the workflow engine, and your conversation history all live locally.
During execution, the data that leaves your computer is what you explicitly choose to send: conversations with a cloud AI provider, and whatever external services your n8n workflows connect to — both entirely under your control. Prosponsive also uses authentication for licensing and offers optional, opt-in analytics to help improve the product — you choose during setup whether to participate.
All local services communicate over encrypted connections using SSL certificates generated during setup. The connection between Prosponsive and n8n, between n8n and the local database, and between you and the Prosponsive interface are all encrypted. Even though everything runs on your machine, connections are secured as if they were crossing the internet.
Sensitive data — API keys, provider credentials, and secure settings — is encrypted at rest using a machine-specific encryption key.
Prosponsive does not require an internet connection to function (assuming you use a local AI provider like Ollama). If your internet goes down, Prosponsive keeps working.
Prosponsive is designed to be extended, not just configured. You can create custom agents, build custom tools, and connect to thousands of external services.
Create agents tailored to your specific needs. An agent definition includes:
Agents are defined in markdown with YAML frontmatter — readable, version-controllable, and easy to duplicate. You can build a team of specialist agents that mirror how your actual work is organized: one for email, one for project management, one for research, one for reporting.
Build any tool you can imagine using n8n's visual workflow designer. Common patterns include:
Prosponsive is not limited to tools that reach out to external systems. Any system that can make an API call to your machine can send data to your Prosponsive agents. A webhook from GitHub, Stripe, your CRM, or a monitoring system can call the Prosponsive API directly, or trigger an n8n workflow that processes the data and then hands it to an agent. Either way, external events become agent tasks automatically.
Your agents filter out what doesn't matter, take action on what does, and only involve you when needed. A webhook fires when a customer submits a support ticket — your agent triages it, pulls context from your CRM, drafts a response, and notifies you only if escalation is required. A batch process dumps daily sales data — your agent summarizes it and posts the digest to your reporting channel. The data flows in, agents handle it, and you see only the outcomes that need your attention.
Through n8n, Prosponsive can connect to over a thousand external services out of the box. If a service has an API, you can build a tool for it — often without writing a single line of code. And because each tool is a workflow, not a single API call, you can combine multiple integrations into one tool that produces a complete outcome.
For a hands-on guide to building tools, see the Developer Guide.
Prosponsive supports 12 languages across its interface, so you can work in the language most natural to you.
| Language | Language (native) |
|---|---|
| English | English |
| Spanish | Español |
| French | Français |
| German | Deutsch |
| Italian | Italiano |
| Portuguese | Português |
| Indonesian | Bahasa Indonesia |
| Thai | ภาษาไทย |
| Japanese | 日本語 |
| Korean | 한국어 |
| Chinese (Simplified) | 简体中文 |
| Vietnamese | Tiếng Việt |
Change your language in Settings. The entire Prosponsive interface — menus, buttons, labels, system messages — switches to your selected language. Agent conversations will also respond in your chosen language when appropriate, though this depends on the AI model's language capabilities.
Scheduled prompts let you automate recurring AI tasks. Combined with async tool use and auto-approvals, they make Prosponsive an always-on system that works while you do not.
A scheduled prompt sends a message to a specific agent at a set time — daily, weekly, or on whatever schedule you define. The agent processes the prompt using its tools and posts the result to the designated channel, just as if you had typed the message yourself. If the tools involved have auto-approval rules, the entire task runs without any human interaction.
Open Settings and find the Scheduled Prompts section. Each scheduled prompt has:
Scheduled prompts run locally on your machine — they do not depend on any cloud service. If your machine is off at the scheduled time, the prompt will run the next time Prosponsive starts.
| Feature | What it means for you |
|---|---|
| Async Tool Use | Tools run in the background. You get results, not loading spinners. |
| Async Group Chat | Multiple agents in a channel respond and use tools in parallel. Post once — parallel work begins across all agents simultaneously. |
| Channels & Threads | Agents and tools scoped to workspaces. Threads keep tasks contained, collapsible, and movable between channels. |
| Filtering & Bulk Approval | Review and approve many pending actions at once — built for task management, not sequential chat. |
| Auto-Approvals | Condition-based rules let trusted tools run without interrupting you. |
| Agent Autonomy | Agents manage threads, archive completed work, and organize your channels. Performance visibility shows you where to optimize. |
| Notifications & Definitions | Actionable notifications, editable agent definitions, and scheduled prompts form a complete always-on control layer. |
| Higher-Value Tools | n8n workflows that orchestrate multiple systems per tool call. One tool orchestrates what would otherwise take multiple separate calls. |
| Provider Agnostic | 7+ providers, switch anytime. Design that minimizes token cost instead of maximizing it. |
| Privacy & Security | Local execution. Encrypted connections. Enterprise security scrutiny. |
| Extensibility | Custom agents, custom workflows, 1000+ integrations via n8n. |
| Multi-Language | 12 languages across the full interface. |
| Scheduled Prompts | Automated recurring tasks. Combined with auto-approvals, Prosponsive works while you don't. |
— Personal Autonomous Agent Platform