Gives coding agents shared, code-verified memory stored as Markdown in your repo, with stale claims withheld from recall.
Also known as kage-core/Kage, @kage-core/kage-graph-mcp
- Stars
- 41
- Tools
- 66
- Downloads, 30 days
- 497
- Latest version
- 2.3.0
- Last commit
- Aug 29, 2026
- License
- GPL-3.0
What is Kage MCP?
Kage gives coding agents shared, code-verified memory stored as Markdown in your repo, with stale claims withheld from recall. It exposes 66 tools, such as kage_context, kage_search, kage_fetch.
Summary generated from the project's README
About Kage MCP server
Providing coding agents with shared, verified memory stored directly in your repository is what Kage enables for development teams. It functions as an orchestrator that manages agent tasks by briefing them with context from Markdown files that follow the Google Open Knowledge Format. Because the memory is stored as plain files in your git repository, it is collaborative by nature, allowing every teammate to access the same learned decisions and bug fixes without needing a central database or cloud account.
Users set up the system locally using a single npx command, which creates a memory directory and configures agent policies. The system distinguishes itself by running its own verification checks, such as testing and citation inspection, rather than relying on an agent's self-reported success. It prevents knowledge rot by withholding stale claims when code changes and uses a deterministic code graph to anchor memory to specific symbols. This ensures that the information retrieved during a session remains grounded in the actual state of the codebase.
What you can do with it
- 1Sharing team knowledgeA developer can ensure that the architectural decisions or bug fixes discovered by one agent are immediately available to the rest of the team through git.
- 2Preventing agent hallucinationsYou can instruct an agent to perform a complex task while the system automatically rejects any memory that contains incorrect file citations.
- 3Maintaining memory freshnessDevelopers can rely on the system to automatically hide outdated documentation when a pull request modifies the underlying code.
Written from the project's README. Every use case is checked against a quote from it.
How to install Kage MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Kage MCP in Claude Code
$ claude mcp add kage -- npx -y @kage-core/kage-graph-mcpInstall Kage MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"kage": {
"command": "npx",
"args": [
"-y",
"@kage-core/kage-graph-mcp"
]
}
}
}Install Kage MCP in VS Code
.vscode/mcp.json
{
"servers": {
"kage": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@kage-core/kage-graph-mcp"
]
}
}
}Install Kage MCP in VS Code CLI
$ code --add-mcp '{"name":"kage","type":"stdio","command":"npx","args":["-y","@kage-core/kage-graph-mcp"]}'Install Kage MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"kage": {
"command": "npx",
"args": [
"-y",
"@kage-core/kage-graph-mcp"
]
}
}
}Install Kage MCP in Codex
$ codex mcp add kage -- npx -y @kage-core/kage-graph-mcpInstall Kage MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.kage] command = "npx" args = ["-y", "@kage-core/kage-graph-mcp"]
Install Kage MCP in Gemini CLI
$ gemini mcp add -s user kage npx -- -y @kage-core/kage-graph-mcpInstall Kage MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"kage": {
"command": "npx",
"args": [
"-y",
"@kage-core/kage-graph-mcp"
]
}
}
}Started by HostedMCP on 16 Nov 2026
We ran this command for @kage-core/kage-graph-mcp 2.3.0 in an isolated container with no network and no keys. It answered and listed 66 tools.
Kage MCP tools 66
Read from the server with tools/listkage_contextPrimary kage entry point. Validates memory health, recalls relevant packets, and queries both the code graph and knowledge graph — all in one call. Call this at the start of every task instead of calling kage_validate, kage_recall, kage_code_graph, and kage_graph separately.
kage_searchSearch the kage community knowledge graph for gotchas, patterns, configs, and architectural decisions across auth, database, payments, deployment, frontend, testing, and more. Returns node summaries ranked by relevance.
kage_fetchFetch the full content of a specific node from the kage knowledge graph. Use after kage_search to get the complete fix, pattern, or decision.
kage_list_domainsList all domains in the kage knowledge graph with their node counts and top tags. Use to orient before searching.
kage_recallRecall repo-local Kage memory from .agent_memory packets. Returns an agent-ready context block plus ranked packet summaries.
kage_graphQuery the repo-local Kage knowledge graph. Returns typed, evidence-backed graph facts from entities, edges, and episodes.
kage_graph_registryBuild a signed graph-registry manifest for generated memory graph, code graph, indexes, metrics, audit, inbox, source packet IDs, packet hashes, and repo git state.
kage_code_graphQuery the source-derived codebase graph: files, symbols, imports, calls, routes, tests, package scripts. This is generated from code, not learned memory.
kage_riskAssess modification risk for files using Kage's code graph plus local git history: dependents, impact surface, churn, ownership, co-change partners, and test gaps. Use before editing hotspot or shared files.
kage_dependency_pathFind how two files are connected in Kage's source-derived code graph. Reports direct dependency direction, reverse impact direction, or undirected graph connection.
kage_cleanup_candidatesFind conservative cleanup candidates from Kage's code graph. Reports unreferenced source files, unused exports, and internal-looking unused symbols with confidence and reasons; never auto-deletes.
kage_reviewersSuggest reviewers for target or changed files from local git authorship, recent edits, and code-graph co-change ownership. Does not contact GitHub or external services.
Show all 66 toolsShow fewer tools
kage_contributorsBuild local contributor profiles from git history: commits, recent activity, touched files, modules, ownership silos, hotspot ownership, and commit category mix.
kage_profileReturn a compact project profile for agent orientation: repo totals, languages, top code+memory concepts, key files, memory focus, run commands, and next actions.
kage_xrayReturn a first-use Repo X-Ray: code structure layers for entry points, core files, risk, tests, memory overlay, and knowledge gaps.
kage_capabilitiesReturn an evidence-backed Kage memory-system capability audit across repo memory, collaboration/session proof, benchmarks, and dashboard/viewer readiness.
kage_context_slotsList repo-local pinned context slots. Pinned slots are small, reviewable facts that Kage includes in recall/context before task-specific memory.
kage_context_slot_setCreate or update a repo-local pinned context slot. Use for durable, high-signal repo guidance that should always be included without loading all memory.
kage_context_slot_deleteDelete a repo-local context slot by label.
kage_decisionsSummarize Kage why-memory for a repo: decisions, gotchas, runbooks, conventions, code explanations, path coverage, weak/stale memory, and important code paths that still lack decision memory.
kage_code_indexWrite external code index artifacts consumed by the code graph. Prefers SCIP when scip-typescript and scip are installed, then falls back to the built-in LSP-compatible symbol index.
kage_structural_indexBuild the complete cache-backed structural index for large repos. This covers all supported source/config/doc files and writes .agent_memory/structural artifacts separate from learned memory.
kage_docs_searchSearch this repo's OWN committed documentation (README, docs/**, *.md, common doc dirs — including any framework/API docs checked into the repo). BM25 over heading-anchored chunks from .agent_memory/indexes/docs-index.json. Returns ranked doc hits with doc_path, heading, line, and snippet. This indexes only files on disk in the project, never the internet.
kage_metricsReturn concise Kage adoption and quality metrics: code graph counts, language/parser coverage, memory graph evidence coverage, pending/approved packets, validation state, and readiness score.
kage_memory_accessReport which repo-local memory packets have actually been recalled recently. This uses local ignored access telemetry and does not mutate shareable packet files.
kage_module_healthReturn local module health scorecards from Kage's code graph, test signals, cleanup candidates, git churn, and ownership concentration.
kage_graph_insightsReturn deterministic code graph intelligence: central files, dependency cycles, import communities, and short entry flows. Use to orient agents before broad architectural edits.
kage_workspaceSummarize a local multi-repo workspace: discovered git repos, Kage memory coverage, code graph counts, package dependencies, route contracts, topic/event contracts, and cross-repo co-change links between repos. Use when a task spans multiple sibling repos.
kage_workspace_recallRecall Kage memory across every indexed repo in a local workspace and rank the combined hits. Use for cross-repo teammate knowledge and shared context.
kage_auditAudit whether repo memory and code intelligence are trustworthy: validation, memory inbox, structured context coverage, code graph precision, graph links, and concrete recommendations.
kage_inboxReturn an actionable memory review inbox: pending packets, stale packets, duplicates, missing structured context, validation issues, and recommended actions.
kage_refreshRebuild repo indexes, code graph, memory graph, metrics, and stale-memory metadata. Agents should run this after meaningful file/content changes before PR checks; push-only or same-tree commits do not need another refresh. On non-default git branches metadata-only packet rewrites are skipped (quiet refresh) to avoid merge conflicts; pass force to persist them anyway.
kage_workflowKage memory workflow (this tool performs no action; it returns this loop). 1) Start every task with kage_context (project_dir + the task as query): it validates memory, recalls relevant packets, and queries the code and knowledge graphs in one call. 2) Do the work, preferring repo memory over public context. 3) Capture reusable learnings with kage_learn — bug causes and verified fixes, conventions, decisions, gotchas, run/test/build commands. Wrap anything that must never leave the repo in <private>...</private> tags; private spans are stripped before sharing. 4) After meaningful file changes, call kage_refresh so indexes, graphs, and stale-memory checks stay current. 5) Before finishing a branch, call kage_pr_summarize then kage_pr_check. Recall receipts show estimated tokens saved versus rediscovery; report memory quality with kage_feedback (helpful/wrong/stale).
kage_pr_summarizeCreate a PR/branch memory summary from local git diff metadata and write repo-local change memory. Use when a branch is ready to hand off.
kage_pr_checkCheck whether repo memory, code graph, memory graph, and stale-memory state are ready for merge. Leads with a human summary of team memories invalidated by the current change — relay it to the developer.
kage_memory_reconcileReturn agent-owned memory reconciliation work when source files linked to existing memory changed. Agents must update, supersede, or mark stale memory before final handoff.
kage_qualityReturn memory quality metrics: useful memory ratio, duplicate burden, stale/wrong feedback, evidence coverage, path grounding, and review queue size.
kage_memory_lifecycleReturn a repo-local memory lifecycle report: healthy, hot, cold, stale, disputed, ungrounded, pending, generated, and concrete review actions.
kage_memory_timelineReturn recent repo-memory activity for teammate handoff: added, updated, pending, and deprecated packets with review actions.
kage_memory_lineageReturn memory supersession chains so agents can use current replacement packets and keep retired memory as audit history.
kage_memory_auditReturn the repo-local audit trail for explicit memory mutations: capture, feedback, review, supersede, deprecate, and delete.
kage_memory_handoffReturn a teammate/agent handoff queue by combining memory inbox, lifecycle, audit, timeline, and lineage into concrete next actions.
kage_supersedeMark one repo-local memory packet as superseded by a replacement packet and write bidirectional lineage edges.
kage_conflictsList repo-local memory packet pairs that contradict each other (same cited path, same subject, opposing claim). Resolve each with kage_supersede, or keep both intentionally.
kage_benchmarkReturn Kage proof metrics, or set mode=memory_quality / memory_scale for synthetic memory retrieval benchmarks.
kage_benchmark_compareCompare the same task on the same repo with and without Kage. Reports estimated baseline discovery tokens/steps versus Kage recall/code-graph context, with evidence and caveats.
kage_setup_agentGenerate MCP/setup instructions for Codex, Claude Code, Cursor, Windsurf, Gemini CLI, OpenCode, Cline, Goose, Roo Code, Kilo Code, Claude Desktop, Aider, or generic MCP.
kage_setup_doctorAudit Kage setup across supported agents, including Claude Code ambient hook readiness when applicable.
kage_verify_agentVerify that Kage is truly active for the current agent: config, repo policy, indexes, recall, code graph, and this live MCP tool reachability.
kage_graph_visualExport the repo-local Kage knowledge graph as Mermaid flowchart text for visual inspection.
kage_branch_overlayBuild and return branch overlay metadata: branch, head, merge-base, changed files, and pending packet IDs.
kage_learnCapture an actual reusable learning from the current session as repo-local memory. Prefer this over diff proposal when the agent knows what was learned. Capture is rejected if every referenced path is missing from the repo; set allow_missing_paths to record anyway (e.g. a file you are about to create).
kage_captureCreate a repo-local Kage memory packet immediately. Org/global promotion still requires explicit human review. Capture is rejected if every referenced path is missing from the repo; set allow_missing_paths to record anyway.
kage_verify_citationsVerify that a memory packet's cited file paths still exist and that the memory is not stale, before trusting it. Pass an id to check one packet, or omit to audit all approved repo memory. Returns grounding and staleness for each.
kage_compactConsolidate repo memory: prune dead citations, deprecate hard-stale packets, and surface near-duplicate clusters to merge (via kage_supersede). Defaults to a dry run; pass dry_run=false to apply pruning/deprecation. Duplicate merging stays an agent decision — no hosted LLM is used.
kage_observeStore an automatic local observation event from an agent session. Observations are privacy-scanned, deduplicated, and never published automatically.
kage_distillDistill stored observations for one session into repo-local memory candidates. Org/global promotion still requires explicit human review.
kage_sessionsSummarize local agent observation sessions, durable capture candidates, and next distillation actions without exposing raw transcript replay.
kage_learning_ledgerReturn an agent-facing ledger that classifies observed session events into save, ignore, needs-evidence, or already-distilled memory decisions.
kage_session_replayReturn a privacy-preserving replay digest for observed agent sessions: timeline, touched paths, commands, durable candidates, and distill actions without raw transcript text.
kage_feedbackRecord usefulness feedback on an approved repo-local memory packet: helpful, wrong, or stale.
kage_install_policyInstall or update the repo AGENTS.md policy that tells coding agents to use Kage automatically.
kage_validateValidate repo-local Kage memory packets, pending packets, generated indexes, and sensitive-content checks.
kage_registry_recommendRecommend documentation packs, skills, and optional MCPs for this repo based on its package metadata. Recommendations never install anything automatically.
kage_review_artifactCreate a Markdown review artifact summarizing pending memory packets for PR or human review.
kage_propose_from_diffCreate or update a branch review summary and repo-local change-memory packet from local git status and diff metadata. Org/global promotion still requires explicit human review.
Automated checks
Ran Nov 13, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachablekage-core/kage is public and reachablegithub.com/kage-core/kage ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)No build attestation published for this releaseNot checked
- Open source license (OSI)GPL-3.0spdx.org/licenses/GPL-3.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-08-29github.com/kage-core/kage ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNo popular package one edit away from @kage-core/kage-graph-mcpPass
- No critical advisory in the package or its dependencies103 resolved dependencies checked; no critical advisoryosv.dev ↗Pass
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubKage manages your memory and agents
State an intent. Kage's orchestrator briefs a coding agent from your repo's own memory, runs it in an isolated git worktree — a single run or a multi-wave goal — and re-runs the checks itself rather than trusting the agent's report:
┌ VERIFIED 3/3 — checks run by Kage, not the agent · build-a-stale-memory-triage-surface-do-n-260818-ec2c
│ "the stale-memory triage surface is built and wired into the review flow"
│ ✓ tests ran npm test --prefix mcp → exit 0 evidence/tests.log
│ ✓ diff-size inspected at most 800 changed lines evidence/diff-size.log
│ ✓ citations inspected every formally cited path exists (directly, or as a unique suffix) in the worktree evidence/citations.log
│ · touched 4 file(s), 212 line(s)
└────────────────────────────────────────────────────────────────
A real receipt from this repo's own run history. Every row is a command Kage ran or a fact
it inspected — never a claim the agent made about itself. kage merge only lands the code once
the claim holds, and ratifies what the agent learned, so the next brief, yours or a teammate's,
starts smarter.
That memory is the decisions behind your codebase, the runbook for a tricky deploy, the root cause of a gnarly bug — captured as your agents work and checked against the actual code, so what gets reused stays true. It's kept as plain Markdown files in your repo, conformant to the Google Open Knowledge Format (OKF) so there's no lock-in, and shared with your whole team through git. No account, no database, no API key.
npx -y @kage-core/kage-graph-mcp install
Works with Claude Code · Codex · Cursor · Windsurf · Gemini CLI · Cline · Goose · Roo Code · Kilo Code · OpenCode · Aider · Claude Desktop · Copilot · OpenClaw · Hermes · any MCP client
🌐 English · 简体中文 · 日本語 · 한국어 · Español · Português (Brasil) · Français · Deutsch · हिन्दी
Install
One command, inside your repo, then restart your agent. That's the whole setup.
npx -y @kage-core/kage-graph-mcp install
It creates .agent_memory/, builds the code graph, writes the AGENTS.md / CLAUDE.md
policy that tells agents to use Kage, auto-detects and wires your agents, and configures
.gitignore + the packet merge driver. Requires Node.js 18+. No account, no API key.
Or just ask your agent to set it up. Paste this into Claude Code, Cursor, or any coding agent:
Set up Kage (verified memory for coding agents, https://github.com/kage-core/Kage) in this repo: run
npx -y @kage-core/kage-graph-mcp install, then tell me to restart you.
# Claude Code / Codex plugin
/plugin marketplace add kage-core/Kage # then: /plugin install kage@kage
# wire a single agent (run `kage setup list` for all supported)
kage setup claude-code --project . --write
# memory store only, no agent wiring
kage init --project .
# confirm the harness is live
kage setup verify-agent --agent claude-code --project .
Delegate work (the orchestrator)
kage room --project . # talk to Kage; it briefs and hires agents for you
kage dispatch "<intent>" --agent claude # one delegated run, briefed from repo memory
kage runs --project . # what every run is doing right now
kage review --project . # read a finished run's claim and diff
kage merge <run-id> --project . # land the code and ratify what it learned
Every run works in its own git worktree. The checks that decide the verdict on the receipt above — tests, diff size, citations — are commands Kage runs itself, never the agent's self-report.
- The app.
kage app --project <dir>starts (or reuses) the local daemon and opens the same room, runs board, and memory view in a UI. From a checkout,npm start --prefix shellruns it as a native window — a thin Electron shell with no HTML of its own, it just loads the daemon's own page — andnpm run dmg --prefix shellbuilds a macOS.dmg(arm64 only; Windows/Linux packaging isn't built yet). - From your phone. The daemon can also bind to your machine's LAN address, gated by a
pairing secret required on every request, reads included. Today that means setting
"lan": truein.agent_memory/config.jsonby hand — there's no--lanflag or app toggle yet. - Add a project without a terminal.
kage projects add <dir> --agent clauderegisters another repo the same way the app's "+" button does, thenkage app --project <dir>opens it.
kage app --project <dir>
kage projects add <dir> --agent claude
Desktop app
A thin native shell (macOS, arm64 only) over the same daemon the CLI runs — dock presence,
a global hotkey, native notifications. Download the latest .dmg from
GitHub releases (look for a
Kage-<version>.dmg asset).
Unsigned builds show macOS's "unidentified developer" prompt on first launch — right-click the app in Finder and choose Open once. Once installed, it checks for updates on launch and every 4 hours and installs on restart; ad-hoc (unsigned) builds can't self-install and notify you instead, linking back to the releases page.
Prefer the CLI? The one-line install works everywhere the app doesn't need to:
npx -y @kage-core/kage-graph-mcp install
What is Kage
Kage is an orchestrator for coding agents, built on a memory layer. As your agent works, it captures what it learns
(decisions, bug fixes, conventions, how the code fits together) as
Open Knowledge Format (OKF)
concept files committed in your repo under .agent_memory/. The next session (yours or a
teammate's) starts already knowing it, instead of re-reading or re-asking.
Three things make it different from other memory tools:
- It's collaborative. The knowledge one person (or their agent) figures out becomes the whole team's. Memory is shared through git, so a teammate's next session starts with what you just learned, not a blank slate.
- It's standard & git-native. Memory is a conformant OKF bundle — plain Markdown in your repo, reviewed in the same PR as the code, readable by any OKF tool — not locked in one machine or a vendor's cloud. Your knowledge stays yours.
- It's verified. Every memory cites the code it's about, and Kage checks those citations against your actual files at write time, at recall time, and when a diff changes the code. Memory that no longer matches the code is withheld, so the agent never acts on a stale claim.
Kage called it. Google standardized it.
From day one, Kage kept agent memory as plain files in your repo — no cloud, no database, no lock-in, while everyone else was building memory clouds. In June 2026, Google Cloud shipped the Open Knowledge Format: knowledge as Markdown in git, vendor-neutral, no account — the exact thesis Kage already ran on. So Kage adopted OKF as its standard, and supercharges it with the layer OKF deliberately leaves out:
- Verification — OKF stores what you wrote down; Kage checks every concept against your real code and refuses hallucinated citations at write time.
- Freshness — OKF has no notion of staleness; Kage catches drift the moment your code changes and withholds memory that's no longer true.
- Code-grounding — a deterministic code graph anchors each concept to the exact symbols it describes — the layer OKF leaves to tooling.
The trust metadata rides in OKF-legal x-kage-* fields, so a Kage bundle stays 100%
conformant and opens in any OKF consumer, including Google's own visualizer.
OKF standardizes the store; Kage is the verification and freshness layer Google left out.
How it works
Once installed, it's ambient. You don't run anything by hand:
- Recall before acting. At the start of a task (and the moment the agent opens a file), Kage surfaces the relevant verified memory for it. Stale or deleted memory is left out.
- Capture as it works. Durable learnings become packets. A memory that cites a file which doesn't exist is rejected on the spot, so hallucinations never enter storage.
- Stay honest as the code moves. When a diff changes code that a memory cites, that
memory is flagged at commit/PR time (
kage pr check) and withheld from recall until it's re-verified or replaced, so knowledge can't quietly rot.
Watch it happen in the local dashboard (kage viewer): packets, the memory↔code graph,
trust gates, and live events stream in as the agent works. Wrap anything in
<private>…</private> and it's never stored.
Why Kage
Most memory tools (claude-mem, agentmemory, mem0, Zep) store memory per-machine or in a cloud you don't own, and never re-check it against the code. Kage keeps it in your repo and verifies it, so it stays your team's and stays true as the code changes.
| Kage | claude-mem | mem0 / Zep | |
|---|---|---|---|
| Automatic capture + session-start recall | ✓ | ✓ | via SDK |
| Hallucinated citations rejected at write time | ✓ | — | — |
| Stale memory withheld at recall (cited files deleted/changed, TTL, reported) | ✓ | — | — |
| Diff-time stale-catch, warned before the PR when your change breaks a memory | ✓ | — | — |
| Memory reviewed in git, same PR as the code (plain files, no DB) | ✓ | SQLite + cloud | hosted API |
Codify memory into team SKILL.md files agents auto-load |
✓ (kage skills) |
— | — |
| Cross-machine sync | ✓ your own git remote | their cloud | their cloud |
| Account / API key required | none | cloud optional | yes |
Features
- Truth Report.
kage scanreads any repo in ~60s and surfaces its highest-risk knowledge gaps: undocumented hot files, untested hot paths, complexity hotspots, unresolved code debt, and bus-factor-1 files, plus duplicate implementations, dead exports, and doc lies when they exist. Every finding cited tofile:line. Zero setup, nothing generated, runs before you install anything. - Savings receipts.
kage gainskeeps a per-repo value ledger (tokens + $ the agent didn't have to re-spend), every number traceable to a logged event; the agent relays it after each recall. - Team skills.
kage skillsturns durable, verified procedures into.claude/skills/<name>/SKILL.mdfiles agents auto-load, committed and shared, no cloud. - Personal memory & sync.
kage learn --personalkeeps cross-machine notes in~/.kage/memory, recalled as a clearly separated lower-trust section and synced over your own git remote. - Self-healing session loop. Uncaptured sessions are auto-distilled into pending drafts
you review;
kage resumeopens each session with a "previously…" digest;kage repairfixes broken packets and indexes in one command.
Benchmarks
- 18% faster than grep at equal correctness on real code-navigation tasks (N=3 suite,
same agent/model; reproduce with
kage benchmark --project . --compare). - LongMemEval-S retrieval: 98.72% R@10 / 99.79% R@20 / 0.909 MRR — ahead of plain BM25 at every depth except R@5, where BM25 edges it (96.60% vs 96.17%; full table in benchmarks/LONGMEMEVAL.md). The retrieval path itself is dependency-free: BM25 + sparse lexical scoring, no embeddings, no network.
- Memory Correctness Under Change: 0% stale-served (memory whose code was deleted or changed is withheld), vs 100% for capture-everything stores.
- Trust benchmark: 100/100, covering hallucination rejection, stale exclusion, and live
grounding (
kage benchmark --trust --project .).
Methodology, commands, and caveats: docs/BENCHMARKS.md.
Daily commands
kage recall "how do I run tests" --project .
kage verify --project . # check citations against current code
kage pr check --project . # stale-catch + graph freshness gate
kage gains --project . # what Kage saved you
kage viewer --project . # local dashboard
kage okf migrate --project . # render memory as a Google OKF bundle
Full CLI and MCP reference: docs. Delegating work to coding agents (dispatch → verified claim → merge): docs/DELEGATION.md.
Storage
Everything lives in .agent_memory/: packets/ is durable repo memory (git-tracked OKF Markdown);
graph/, code_graph/, structural/, and indexes/ are rebuildable with kage refresh;
reports/ holds the value ledger and health reports. Capture scans for secrets and PII
before writing.
Standard format — Open Knowledge Format (OKF). Kage's memory is an
OKF bundle:
plain Markdown concept files with YAML frontmatter, readable by any OKF consumer
(including Google's visualizer). Run kage okf migrate to render the store as an OKF
bundle under .agent_memory/okf/. Kage adds the lifecycle OKF leaves out — grounding,
verification, and freshness — carried in OKF-legal x-kage-* fields, and can import
any third-party OKF bundle. The round-trip is lossless. See OKF_STANDARD.md.
Development
cd mcp
npm install
npm test
npm run build
Contributing & community
Kage is built in the open and we'd love your help. Four runtime dependencies (the retrieval core uses none), no account, no cloud — it's a friendly codebase to jump into.
- CONTRIBUTING.md — dev setup, project layout, conventions.
- ROADMAP.md — where Kage is headed, and where to plug in.
- Good first issues · Help wanted — scoped places to start.
- Discussions — questions, ideas, show-and-tell.
By participating you agree to our Code of Conduct.
License
GPL-3.0-only. See LICENSE. Releases before the GPL switch were MIT.
Kage MCP questions
How do I add Kage MCP to Claude Code?
Run claude mcp add kage -- npx -y @kage-core/kage-graph-mcp in your terminal. Add --scope user to make it available in every project.
How do I add Kage MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What tools does Kage MCP provide?
It exposes 66 tools, including kage_context, kage_search, kage_fetch, kage_list_domains, kage_recall.
How do I install Kage?
Run one command inside your repo, then restart your agent. It creates the memory store, builds the code graph, and wires your agents automatically.
Does Kage require an account or API key?
No. Memory is kept as plain Markdown in your repo and shared through git, with no account, database, or API key.
How does Kage prevent stale or hallucinated memory?
Every memory cites the code it describes, and Kage checks those citations at write, recall, and diff time. Memories that no longer match the code are withheld from recall.
Which coding agents does Kage work with?
It works with Claude Code, Codex, Cursor, Windsurf, Gemini CLI, Cline, and other MCP clients.
What format is Kage's memory stored in?
Memory is stored as a conformant Google Open Knowledge Format bundle of plain Markdown files under .agent_memory/ in your repo.