Get This Tool
Open Kioku
Pricing
- Model
- Free
Summary
Coding agents running grep-and-guess loops burn through context window before they understand the repo — then edit the wrong files and break tests they never knew existed. Open Kioku is the local evidence layer that stops that loop before it starts.
Open Kioku indexes a repository into a local SQLite and Tantivy store, then exposes that index to MCP clients — Claude Code, Cursor, Codex — so agents build structured, evidence-backed plans before touching a single file. The pre-edit plan names exact file ranges, affected symbols, caller graphs, and validation candidates. The post-edit verify step checks that changes stayed inside the declared boundary. No source leaves the machine; write tools stay gated unless you explicitly unlock them. The ceiling arrives when the repo or language falls outside what the local indexer can parse — at that point the graph edges simply are not there, and the agent falls back to inference.
Bottom line: Pick this when you are running an MCP coding agent against a large local codebase and need symbol-accurate plans before edits land — but expect to hit language coverage gaps that force the agent back to text-match inference on unsupported file types.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Local SQLite and Tantivy index with no source upload, so repositories under data-residency constraints or containing unreleased code never leave the machine — the class of hosted-index compliance problems does not arise.
- Symbol-resolution and graph-edge tracing replace text-match inference for impact analysis, which means an agent planning a token-expiration change gets the actual callers and affected modules instead of a list of files that contain the string.
- Pre-edit plans include explicit allowed and forbidden file boundaries, so the agent is told what is out of scope before it starts — reducing the drift where an edit intended for two files touches six.
- Validation candidates are selected before the edit begins rather than after, so the agent runs the right tests against the right change instead of discovering the gap in the CI failure.
- MCP stdio transport connects to Claude Code, Cursor, and Codex without a dedicated plugin per client, so one local index serves multiple agent surfaces from a single install.
Cons
Sign in to edit- Language coverage determines graph completeness: repositories mixing a supported language with an unsupported one produce incomplete impact traces for the unsupported files, and the agent reverts to text-match inference for those paths — which is the same baseline the tool exists to replace. Teams with polyglot monorepos check the language support list before committing to the workflow.
- There is no API and no hosted option, which means teams building CI pipelines or cloud-based agent workflows that cannot run a local process have no path to integrate Open Kioku — those teams evaluate hosted code intelligence services that accept repository uploads instead.
- The index must be rebuilt or re-run after significant repository changes; the docs describe no incremental or watch-mode indexing, so teams with high-commit-velocity repositories that need fresh graph edges before every agent run carry a re-index step as an explicit pipeline cost.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- npm, local CLI
- API Available
- No
- Self-Hosted
- Yes
- Last Updated
- 2026-07-05T08:15:56.645Z
Best For
Who it's for
- AI coding agents using MCP clients
- Large local codebases that cannot be uploaded
- Pre-edit planning and post-edit verification workflows
What it does well
- Plan code changes with indexed evidence before editing
- Trace symbol references and impact across a repository
- Select relevant tests for validation candidates
- Verify bounded edits after changes
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Open Kioku free?
- Yes — Open Kioku is fully free to use. There is no paid tier.
- Is Open Kioku open source?
- Yes. Open Kioku is open source.
- Can I self-host Open Kioku?
- Yes. Open Kioku supports self-hosting on your own infrastructure.
- What platforms does Open Kioku support?
- Open Kioku is available on: npm, local CLI.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Best Open Kioku alternatives →
Curated lists that include this category
Coding agents that navigate unfamiliar codebases with repeated grep and glob calls trade context window for guesswork: references are inferred from text, test selection happens after the edit, and the agent has already committed to the wrong surface. Open Kioku installs locally via npm, indexes a repository once into a local SQLite and Tantivy store, and connects to any MCP-compatible client through stdio transport. The workflow is a fixed loop — index once, then for each change: search ranked evidence, resolve symbols, trace graph impact, produce a structured pre-edit plan with explicit edit boundaries and validation candidates, make the edit, verify the boundary held.
The defining architectural choice is that nothing leaves the machine and nothing requires an embeddings API or hosted index. The vendor states the tool runs read-only by default, with write tools gated behind explicit enablement. Plans are structured markdown that name allowed and forbidden file paths, list validation candidates as runnable test commands, and report evidence quality signals — exact-reference availability, coverage, runtime data — rather than hiding confidence gaps inside a summary.
Open Kioku fits teams whose codebases cannot or should not be uploaded to a hosted service and who are already using an MCP-capable agent. It fills the gap between what grep finds and what an agent needs to know before acting — specifically: which symbols are affected, which callers exist in the graph, and which tests are relevant. The gap that remains is language coverage: the local indexer’s graph edges only exist for languages the parser handles, so repositories mixing supported and unsupported languages will see incomplete impact traces on the unsupported side, and agents fall back to text-match inference for those files.
Installation follows four shell commands: global npm install, `ok init` with an absolute repo path, `ok index`, and `ok mcp install` targeting the client. The printed MCP config snippet pastes into Cursor, Claude Code, or another MCP client. The tool carries no hosted component, no pricing tier, and no API dependency — the index lives on disk alongside the repository.