Overview
Nachalnik is an agent runtime in which the context, the tools, the permissions and the requests are explicit state that can be read, changed and put back. It is described as a loop that is a state machine, a context that is a list of identified values, and an append-only log of everything that happened. The runtime has a handful of dependencies, no unsafe, no network, and no prompt, and is meant to stay boring.
Other crates in the workspace are built on top of it. Kamchatka is a terminal agent built on the runtime. Nachalnik-mcp is a bridge to MCP servers, so that a tool somebody else wrote is a Tool like any other. Nachalnik-eval is a benchmark for model introspection. Nachalnik-providers contains the three dialects — OpenAI chat-completions, Google's own and Anthropic's own — streamed, retried and interruptible, behind one trait. The runtime opens no sockets by design; this is where the sockets are.
Nachalnik is complete for what the runtime claims to cover — the state machine, the context model, permissions, the event stream, sessions, and projection — with every invariant it states held by a test. Deliberately not included, and not planned for the core: MCP, subagents, an editor protocol, a daemon, a CLI, or a prompt library. Those belong on top of it, and that is what the rest of the workspace is for.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Teams use nachalnik to build an agent runtime where context, tools, permissions, and requests are explicit state.
- Teams use nachalnik to run a loop as a state machine with a context list and an append-only event log.
- Teams use nachalnik-mcp to bridge MCP servers so a tool somebody else wrote is a Tool like any other.
- Teams use nachalnik-eval to benchmark model introspection by comparing a model's claim about its own context against a moved copy.
- Teams use nachalnik-providers to connect OpenAI chat-completions, Google's own, and Anthropic's own dialects behind one trait.
- Teams use nachalnik to keep the core free of MCP, subagents, an editor protocol, a daemon, a CLI, or a prompt library.
Features
AI assistant
Natural language interface
Workflow automation
Reviews
No reviews yet. Be the first to share your experience!