ToolCairnMost queries start at either search_tools (single tool) or get_stack (multi-layer stack). From there, the graph of related tools guides deeper actions — version checks, comparisons, and feedback.
Search & Discovery
Find the right tool for a job. Start here for most agent queries.
Stacks & Analysis
Build coordinated stacks and compare picks.
get_stackremoteBuild a multi-layer tool stack with resolved versions.
compare_toolsremoteHead-to-head comparison using health signals + graph relationships.
check_compatibilityremoteVersion-aware compatibility check across 35 package registries.
verify_suggestionremoteValidate agent-suggested tools against the graph.
Diagnosis
Troubleshoot after a tool is in use.
Feedback & Contributions
Close the loop so future recommendations get better.
report_outcomeremoteRecord whether a tool worked out (success / failure / replaced / pending).
feedbackremoteAgent-only channel for problems with ToolCairn itself — wrong, broken, or low-quality responses. Free of daily quota.
suggest_graph_updateremoteSuggest a new tool, edge, or category for the graph.
Auth & Setup
One-time configuration on first run.
Project Config
Maintain the per-project .toolcairn/config.json manifest.
What's new in v0.9.0
- Version-aware
check_compatibility— pass optionaltool_a_version/tool_b_versionto evaluate declared peer ranges across 35 package registries (semver, pep440, maven, composer, ruby, cargo). Falls back to graph edges when version metadata is unavailable. - Cross-stack version resolution in
get_stack— every pick ships withrecommended,latest,registry, and a per-tool status. The response also carries acompatibility_matrixand rollupstack_compatibilityso you can tell at a glance whether the layers fit together. - Keyword-sentence sub-needs in
get_stackdrive precise per-layer keyword search — dramatically higher precision vs. plain-string sub-needs.