Short answers, with links to the page that explains more.
Do I need to create tasks myself?
No. You keep talking to your agent. For substantial work, an agent with the recommended instructions finds an existing task or creates one. The CLI is there when you want to look or script. See Working with your agent.
Does every request become a task?
No. Tasks are for work worth keeping track of: features, bug fixes, refactors, investigations, anything that may span sessions. Questions, code reading, and tiny edits are handled directly.
Do I have to tell the agent to use Istok every time?
Usually not. Once MCP is connected and the agent’s instructions describe the workflow, it uses Istok on its own in registered repositories. See the Quick start.
Can I set Istok up once for all repositories?
Yes. Add istok mcp to your agent at user scope; each server process picks the project of the directory it starts in. You still register each repository once, by asking your agent or running istok init. Agent instructions can be global, per project, or both.
Will the agent register repositories on its own?
That depends on your instructions. The recommended ones tell the agent to use Istok in registered repositories and to register one only when you ask. Where Istok is required, you can commit stricter project instructions next to the code.
Can I use several coding agents on one project?
Yes. Project state is not tied to one agent: Codex can start a task and Claude Code can continue it. Give each agent its own --actor-id so you can tell their runs apart. See Several agents, one project.
What happens when I start a new session?
The conversation starts fresh, but the tasks, progress, runs, validations, and saved context are still there. Ask the agent to continue the work and it reads them instead of you retyping the history.
What happens if a session stops in the middle of a task?
Everything recorded so far stays. The run becomes stale when heartbeats stop; the same agent can recover it, the next claim after the lease expires replaces it, or you can abandon it. See Tasks and runs.
What happens if validation fails?
The failure is recorded in the run, and the task stays open. The agent fixes the problem and validates again. See Validation.
Why both tasks and runs?
A task is what needs to be done; a run is one attempt to do it. The task survives interrupted or failed attempts, and each run keeps the record of its own attempt. See Tasks and runs.
Do I need to understand leases and revisions?
Not for normal use. They matter when you drive Istok from the CLI, troubleshoot interrupted work, or build an integration. See istok task and istok run.
How are agent instructions different from Istok context?
Instruction files such as AGENTS.md or CLAUDE.md tell the agent how to work, for example when to use Istok. Istok context is what the project should remember: decisions, constraints, conventions, notes. See Project rules and knowledge.
How is context different from knowledge?
Context records are short and delivered to runs automatically, always or when relevant. Knowledge items are longer, reviewed documents; runs get only their titles and summaries, and the agent reads the full item when needed. See Context and knowledge.
Do I need to maintain the code index?
No. Istok checks freshness whenever search, graph reads, or a claim need the index. istok index status is for inspection; istok index rebuild is for repairs.
What are MCP profiles?
They control which tools an agent gets. The default, worker, covers normal work. supervisor adds abandoning runs, and admin adds project and deletion management. See MCP tools.
Is there a graphical interface?
Yes. istok ui opens a local web UI with your projects, tasks, runs, validations, and the context each run received. See the Web UI guide.
Why does the web UI have no login?
It listens only on 127.0.0.1, rejects requests from other websites in your browser, and is meant to run as a background service, where a token would get in the way. It does not protect against other programs on the same machine, so do not run it on a computer you share with other people.
Does Istok write files into my repository?
No. Projects, tasks, context, and knowledge live in the Istok database; indexes and artifacts in Istok’s data directory. Knowledge becomes a file only when you export it to a path you choose.
Where does Istok keep its data? Is anything sent anywhere?
Everything is local: a SQLite database, code indexes, and artifacts under your user data directory. There is no account and no cloud service. See Configuration.
Can I use Istok without an agent?
Yes. The CLI covers the same workflow for inspection, scripting, and direct control. Start with the CLI overview.
What if the agent says Istok is unavailable?
Check istok version in a terminal, then the agent’s MCP configuration. An agent that cannot reach Istok should say so rather than guess project state. See Troubleshooting.