- Published on
Show HN: PrimerKit, financial market work at agent speed
- Authors
- Name
- DeepHorus
PrimerKit, an open-source workspace for financial market work, has launched on Show HN. It lets an AI agent handle market data tasks rather than simply generating a confident paragraph. Users can ask questions in plain language and see the exact data lines behind each answer.
The tool connects to AI apps such as Claude Code, Codex, and Gemini CLI, or runs in its own terminal. It also supports hosted models on a user's own key or local models through Ollama and LM Studio. PreReason serves as the default data source. Brokers stay off until the user enables one with /connect.
PrimerKit offers eight commands: /digest, /compare, /brief, /journal, /chart, /model, /catalog, and /connect. /digest reports what moved in rates and markets. /compare checks a briefing against web search. /brief presents a briefing as a board, while /journal keeps every run on the user's computer.
Agents can test a user-written rule against past data. They can run that rule each morning on the computer's schedule. When the user approves, they can place a paper order. This turns market analysis from a one-off chat into a repeatable, auditable workflow.
The project is open source and states that nothing in it is investment advice. The model never proposes a strategy. It works from market data or from rules the user writes. That design may appeal to analysts concerned about AI hallucinations and unclear accountability.
PrimerKit reflects a broader shift toward agentic tools that execute tasks, not just discuss them. By linking data, testing, scheduling, and paper trading, it aims to close the gap from idea to action. Future updates could add more brokers and data sources. For now, the tool is available for users to try.
