OpenLore CLI
A code graph and guardrails for agents: one call shows what a task touches, and a check stops an unsafe change
- 355 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The install command edits the agent configuration, registers an MCP server and indexes the whole repository
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/openloreDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The install command edits the agent configuration, registers an MCP server and indexes the whole repository
This is third-party code. Review the repository files before installing.
What it does
OpenLore runs a one-time static analysis of a repository and keeps a knowledge graph. The graph holds calls, types and tests, plus architectural decisions, infrastructure as code and drift from the spec. The agent queries the graph instead of reading files for a long time. No language model sits between calls, so the same question returns the same answer, and no API key is needed. Memory works through the `orient` call: given a task description it returns the relevant functions with their callers, plus matching specs, tests and insertion points. Notes and decisions anchored to a symbol survive a rename or move. When a cached fact has aged or the repository has moved, OpenLore warns the agent. Guardrails look at what the agent changes. `certify_public_surface` marks each changed export as breaking, non-breaking or potentially breaking and names the affected consumers. `change_impact_certificate` notices when an edit opens a new path into a boundary you declared. Before the agent calls code dead, `verify_claim` answers confirmed, refuted or unverifiable with a citation. `openlore enforce` blocks a commit on findings you marked as blocking.
Who it is for. For teams with large repositories where the agent burns many steps finding the right code and sometimes breaks someone else's callers.
Good fit when
- The agent starts each task by reading a dozen files and you want to hand it a ready map: functions, callers, tests
- Before a commit you need to know which consumers a change to a public export will break
- You want to restrict edits near a declared boundary such as the payments layer
- The agent is about to delete code as dead and you want that checked with a citation
- A repository mixes application code, infrastructure as code and recorded architectural decisions, and you want one query across all three
Not a fit when
- The repository is tiny: the agent will read it whole anyway and an index adds a step
- You run Node.js below 22.19: the package targets 22.19 and newer
- You want natural-language answers about what code means: the graph is deterministic and built by static analysis with no model
- You cannot add anything to the repository: `openlore install` wires the agent and builds the index
Example request
Call orient for the task «add a --since flag to the blast-radius command» and show the functions, their callers and the insertion pointsLimitations
Node.js 22.19 or newer is required. The index is built once at install time and updated afterwards, but facts in it can age. Some capabilities are opt-in: `openlore features` shows which are active and how to enable the rest, such as embeddings, the commit hook and the spec store.
OpenLore supports 21 programming languages and 12 infrastructure-as-code ecosystems. Code in unsupported languages stays out of the graph, and the agent still reads it file by file.
How to disable. Remove the MCP server entry and the wired instructions from your agent configuration, then uninstall the package with `npm uninstall -g openlore`. For removing the index itself, see the command reference in the repository.
FAQ
How do I install OpenLore in Claude Code?
Run `npm install -g openlore && openlore install` at the repository root. The command detects your agent or an AGENTS.md file, wires the orient call, registers the MCP server and builds the index. Besides Claude Code it supports Cursor, Cline, Continue and Pi.
Does it need a language model or an API key?
No. The analysis is static and no model sits between calls, so the same question returns the same answer. Optional capabilities such as embeddings are enabled separately through `openlore features`.
What does an orient answer look like in practice?
On the demo task «add a --since flag to the blast-radius command» it named the computeBlastRadius function and the five places that call it, and ranked the insertion points.
How do I stop a commit with an unsafe change?
`openlore enforce` blocks a commit on findings marked as blocking, on new debt in frozen areas and on an unverifiable frozen baseline. By default it only warns.
How is OpenLore different from plain code search?
It holds a graph of calls, types, tests and decisions and answers questions about the impact of a change: who breaks when an export changes, whether a path into a guarded boundary opened. Text search does not see those links.
Discussion
No questions or reviews yet.
Related
CLI toolsA code-review CLI and plugin for agents: reads a git diff and writes line-level comments with context
A Claude Code plugin with seven specialized subagents and approval gates: the agent proposes a plan and waits for confirmation before touching code
Code graph and project memory in one Rust program: the agent gets excerpts with file and line citations
A CLI that builds a dependency graph from code with no AI model and serves it to agents over MCP: what an edit touches, whether layers hold
OpenCode plugin that turns a session into a team of agents led by an architect and runs the work through plan phases
A Claude Code and Codex plugin with skills for code research, tests, refactoring and independent review of finished changes
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OpenLore (this entry) | no | 355 | ||
| codebase-memory-mcp | no | not required | 47,795 | an MCP server with a code graph and structural queries, with no commit gate |
| CodeGraph | no | not required | 75,918 | a local code graph with entry points and change blast radius |
| GitNexus | no | not required | 48,123 | indexes a repository into a knowledge graph with call chains and change blast radius |
README in short
The description states the problem: agents re-read the same files every task and get no warnings about breaking changes. Next come one-command install, sections on memory and governance, a with-and-without table for the same task, a test summary and a comparison with alternatives.