Agents
Prostir’s AI agent builder starts with the job you describe in Builder. Create an agent that answers from your knowledge, uses the actions you allow, and works through a supported chat or channel. Start with one job you can check from beginning to end.
For the usual MCP path, model-token cost to the Agent creator is $0: the customer's AI client supplies the model under that customer's subscription and limits.
Available
You keep explaining the same context and carrying work between conversations and tools. A useful assistant needs your sources, rules and permissions in one place. A policy file, a customer record and a tool connection each hold part of the answer, while you still coordinate the handoff.
An agent with a clear job, your business context and a working path for the people who need it. You can test the agent with real requests, inspect its sources and allowed actions, and refine it in Builder before inviting others.
Agents
Tell Builder who needs help, which questions or tasks repeat, and what a useful result looks like in your business.
Add the needed sources and approved connections. Test routine questions and cases that should stop or ask you.
Choose access and verify the supported client or channel. Ask Builder to refine the behavior as you learn.
Available
agent
Connect your sources and ideas, then inspect the facts behind an answer.
agent
Keep contacts and other structured records in tables, with clear permission to read or change them.
agent
Choose the tools your agent needs and test their permissions before letting it act.
agent
Use the same configured agent in a supported website widget, messaging channel or AI client.
Follow a real interaction in Prostir. Pause whenever you want a closer look.
A useful first step
A ready Builder brief and a five-point launch check. Use it to define a useful job before you spend time building the wrong assistant.
You can edit this brief in Prostir before starting Builder.
Plan your first step
Choose the Agent, knowledge and operation allowances your work needs. Start with a 7-day Creator trial with no card, then compare the current plans inside Prostir.
Through a compatible external MCP client, the model runs in that client’s environment. The Agent creator does not pay those model tokens; the client’s subscription and limits apply.
Choose AI Credits or your own provider key for supported hosted runs. Paid external tools, seller charges and payment-provider fees remain separate.
No. You describe the behavior, add knowledge, set access and pick channels without code. Code only comes in if you want a custom JavaScript tool or a special integration. Prostir handles the hosting and access.
No. A new agent asks people to sign in by default. Opening a free agent to anyone without sign-in is a separate choice you have to confirm.
For the usual MCP path, model-token cost to the Agent creator is $0. When a customer uses the Agent through a supported MCP connector in ChatGPT or Claude, the model runs in that customer's AI environment under the customer's subscription and limits. The creator pays for the Prostir plan, not those model tokens. AI Credits are needed only when the creator separately chooses a Prostir-hosted model run, such as a supported channel or widget. BYOK is optional and bills the creator's own provider account. Paid external tools remain separate.
Yes. A published Agent gets its own hosted page at https://{slug}.ai.prostir.build and its MCP endpoint at /mcp. Discuss a branded custom domain as a custom-plan scope. It needs domain-ownership, DNS, TLS, routing and client-auth review; standard plans do not promise self-service custom domains.
A chat is useful for a question in front of you. An Agent gives a repeated job a defined set of rules, Knowledge, allowed tools and access. You can use that setup in compatible AI clients, while still checking its answers and actions.
Read the full guide →You can describe the job in Builder and choose supported tools, but a connection still needs the right credentials and permissions. A system without a suitable supported connection may need development. Prostir does not make every API compatible or repair external systems automatically.
Read the full guide →AI can still misunderstand a request, use weak information or fail during an action. Give the Agent good sources and clear limits, then test examples that include mistakes. For a Flow, inspect the run and saved result rather than treating a confident reply as proof of completion.
Read the full guide →Keep tools that already do their job. Prostir adds a place to define a product's Knowledge, rules, allowed actions and access, with a shared Builder. Agents can be used in compatible AI clients. A Prostir Flow is useful when its explicit steps fit the job; migration is not a requirement.
Read the full guide →Decide which people, products and tools may use each piece of information. Give the job only the access it needs and consider where the chosen model processes data. Company is a private workspace; a private-looking link alone is not access control or proof of security certification.
Read the full guide →Available
Pick one customer question or routine task and build an agent for it with your own knowledge. Start your 7-day Creator trial, no card needed.