Conversations let agents build on each other.
A conversation keeps variables and prompts. A follow-up is searched like a new request, and the matching agent continues from that record. No agent calls another; the conversation carries the work from one agent to the next.
What a conversation keeps
When an agent ends or pauses, the platform saves the conversation’s state automatically. The next run in the conversation, whether by the same agent or another, starts from that state. Other conversations cannot see it. A conversation also belongs to one user, so other users cannot see it either.
The runs so far. Which agent ran, what inputs it received, what it answered, and where it stopped if it paused to ask a question.
The prompts. The history of every prompt used in the conversation: the user’s turns, the model’s answers, and the tool results. This history lets a model handling a follow-up remember the earlier exchange.
The variables. The conversation variables that every agent in the conversation can read and write, plus each agent’s own variables. An agent stores information that later agents will need, such as trip details, in a conversation variable.
Prompts carry the exchange
A prompt is what an llm command sends to a model: system text, user text, and the history of the exchange so far. Each prompt has a scope that determines how long its history lasts and which agents share it.
conversation (the default). One history for the whole conversation, shared by every agent that has a prompt with the same name. A second agent can see what was said to the first.
agent. The agent’s own history for the whole conversation. Other agents in the same conversation cannot see it.
local. A history for one run, empty at the start of every run. Use it for one-off tasks, such as classifying or summarizing, where earlier runs should not affect the answer.
Each prompt has a name, and that name connects agents. A second agent continues the first agent’s exchange only when its prompt has the same name and uses the conversation scope. A prompt with a different name starts its own history. Prompts explains how prompts are written and evaluated.
Agree on names
Agents build on each other through names: the prompt name that continues the exchange, and the conversation variables that pass one agent’s findings to the next. The system does not enforce these names, so an enterprise must agree on them before its agents can work together.
A naming scheme can be short: one standard prompt name, such as chat, for every agent’s exchange with the user, and one name for each item agents pass to each other, such as conv.trip, conv.customer, or conv.order, along with what it holds. An agent that follows the scheme fits into any conversation in the account. An agent that invents its own names starts with nothing and leaves nothing the next agent can find. See Agreeing on names in the documentation.
A follow-up can go to a different agent
A follow-up in a conversation is searched like a new request. If it belongs to another agent, that agent runs next with the conversation’s state already in place. If no new agent matches, the agent that answered last answers again.
What an interface does with the matches depends on its search behavior. The UI runs the best match immediately; your own interface can let the user choose between the two best matches.
The client always decides which agent runs next. An agent cannot run another agent. A request with several parts is coordinated from outside: an agent harness with the Tactet skill splits the request, sends each part through search and execAgent in the same conversation, and combines the results into one answer. See How agents build on each other.
Ask for something, then follow up
The search and conversation demo runs in a live Tactet account with 200 agents, from human resources and finance to engineering and sales. Ask for something, then ask a follow-up and watch the conversation move to another agent that knows what the first one found.
Kept safe, kept yours
State is encrypted before it leaves the agent server, and the cloud service never decrypts it. An account can keep the state in its own enterprise instead of storing it in the cloud service. On the free service, a conversation that has been unused for 7 days is deleted unless the user pins it. A user can also delete a conversation at any time.
Where to read more
Conversation state covers what is saved and when. Agents building on each other covers who decides what runs next, and Prompts covers scope and the prompt name. In the REST API, running agents covers the call above and conversations covers listing, renaming, pinning and deleting.

