Status basis: npm_downloads_7d momentum fell to -39 per 3d from 29 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -39 per 3d from 29 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -26 per 3d from 23 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -16 per 3d from 13 (<= 40% of the prior window)
- Accelerating npm_downloads_7d gained 29 in the last 3d vs 2 in the 3d before (ratio 14.5x, threshold 1.5x)
- Cooling npm_downloads_7d momentum fell to -16 per 3d from 13 (<= 40% of the prior window)
- Accelerating npm_downloads_7d gained 23 in the last 3d vs 2 in the 3d before (ratio 11.5x, threshold 1.5x)
- Cooling npm_downloads_7d momentum fell to 13 per 3d from 45 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to 2 per 3d from 21 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to 2 per 3d from 21 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -960 per 3d from 599 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -949 per 3d from 599 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -949 per 3d from 599 (<= 40% of the prior window)
- first observed Kanon started keeping permanent history for this agent
| Machine-callable | — | unknown | — |
| Open source | 1 | inferred | 1 quote(s) |
| Self-hostable | 1 | inferred | 1 quote(s) |
| Bring your own key | — | unknown | — |
| Autonomy level | — | unknown | — |
| Pricing model | — | unknown | — |
| Integrations | — | unknown | — |
“Unknown” means we have not verified it — it is not a “no”. Hard filters never treat unknown as false.
Its combination of a local code graph and a refusal to guess addresses a core reliability challenge for AI agents in complex code environments.
Evidence quotesverbatim, from the product’s own materials
“public GitHub repository (README fetched)”— structural
“A local code-context engine for AI coding agents. Sonde indexes a TypeScript, Python, or Swift repository into a symbol-level graph in SQLite”— readme
“exposes three MCP tools — `find_symbols`, `query_graph`, and `get_impact_radius`”— readme
“registers sonde as an MCP server in this project's `.mcp.json`”— readme
FAQ
What is Sonde, a local code graph for AI agents that refuses to guess?
Sonde is a local code graph tool that prevents AI agents from making guesses, improving their accuracy when understanding and working with code.
What does Sonde, a local code graph for AI agents that refuses to guess do?
Enable AI agents to accurately understand and act on code context within a local codebase without relying on speculative inference.
Why does Sonde, a local code graph for AI agents that refuses to guess matter?
Its combination of a local code graph and a refusal to guess addresses a core reliability challenge for AI agents in complex code environments.
Is Sonde, a local code graph for AI agents that refuses to guess open source?
Yes — Sonde, a local code graph for AI agents that refuses to guess is open source.
Can I self-host Sonde, a local code graph for AI agents that refuses to guess?
Yes — Sonde, a local code graph for AI agents that refuses to guess can be self-hosted.
How popular is Sonde, a local code graph for AI agents that refuses to guess?
As tracked by KanonAgent: 2 upvotes (first indexed 2026-08-31).
What are the best Sonde, a local code graph for AI agents that refuses to guess alternatives?
Similar AI agents tracked by KanonAgent: Stealth Venture, taste-skill, open-design, deer-flow, ruflo, career-ops.