Kontexta: workspace + context engine for coding agentsREV 7.2.2

Your agents have memory. It just isn't yours.

Every coding agent keeps its own silo. kontexta is one local workspace they all share: searchable, versioned, portable.

  • Local-first
  • Optional E2EE sync
  • Your machine, your data
  • Apache-2.0
$npx kontexta start
FIG. 0: LIVE LOOP

Click any stage above to jump in. Claude Code fixes the bug and journals it; Hermes recalls it from the same local vault. Wire up your clients →

kontexta MCP server score
SEC. 01: FIELD DEMONSTRATION

Product in Action


FULL WALKTHROUGH · 04:50 MIN

Full Product Walkthrough

See Kontexta register projects, run sandboxed commands, distill journals, detect calendar conflicts, publish docs, and deliver cross-project intelligence - all in a single session.

Interactive Dashboard: Local-first dashboard for managing memory, browsing context, and inspecting the Brain live.

Interactive Dashboard

Local-first dashboard for managing memory, browsing context, and inspecting the Brain live.

Coding Agents & MCP: Auto-configure the kxta MCP server and native hooks across Claude Code, Cursor, Gemini, and more.

Coding Agents & MCP

Auto-configure the kxta MCP server and native hooks across Claude Code, Cursor, Gemini, and more.

Agent Journal & Distillation: Live stream of tool calls and shell actions, distilled into persistent topic notes.

Agent Journal & Distillation

Live stream of tool calls and shell actions, distilled into persistent topic notes.

Governed Hands Execution: Define project-specific tools with approval gates, single-use tokens, and schema verification.

Governed Hands Execution

Define project-specific tools with approval gates, single-use tokens, and schema verification.

SEC. 02: MEASURED · 2026-10-01 · COMMIT fe5177a

Context Efficiency, Measured


Dumping entire codebases into your AI's context window is expensive and degrades response quality. Kontexta's MCP server retrieves only the files a task needs. Here is the measurement, the method and its limits.

66.8%
Fewer file tokens loaded per task
3.0×
Less context to reach the same files
41,171
Tokens saved across 4 scenarios
SEC. 03d: RANKING

Search that ranks, not just matches


Keyword search finds candidates. A local neural model puts the best one first, and every result shows its score.

On-device
A small cross-encoder reranks keyword hits. No API call, no key.
Bundled
The model ships in the npm package. Nothing is downloaded at runtime.
Fails soft
If the 300 ms budget runs out you get keyword order, flagged, never an error.

See how ranking works, with measured latency and limits →

SEC. 03: SYSTEM SCHEMATIC

Brain → Hands → Eyes


A closed feedback loop that turns observation into durable knowledge.

Brain

The context engine.

Sealed SQLite vault with FTS5 + local cross-encoder search, surgical section edits, built-in revision history, and E2E-encrypted multi-machine sync. Plus calendar, tags, publishing, and portable context bundles.

See what's inside

Hands

Governed execution.

Sandboxed, project-defined commands declared in kontexta.json. Approval-gated when destructive. The only memory that can act.

See how Hands works

Eyes

Memory that edits itself.

Every MCP tool call is captured to a per-project event log without agent involvement. Distillation turns raw events into per-topic markdown the next session searches instead of re-deriving.

See the distill loop
ACTIVE SUBSYSTEM · BRAIN
STORAGE
Sealed SQLite + Brotli revisions (zero repo clutter)
RETRIEVAL
FTS5 + local cross-encoder reranking & token budgets
PORTABILITY
AES-256-GCM Brain sync + kontexta export / import
●files_search(query: "auth", kind: "dictionary") → est_tokens: 140Inspect Brain sheet →
SEC. 03e: SYNC ACROSS MACHINES (OPTIONAL)

One brain, every machine.


Local-first stays the default. Add a relay you host and your brain follows you, with a copy that survives a lost laptop. The relay can't read any of it.

Blind by design

End-to-end encrypted.

Payloads are encrypted on your machine with AES-256-GCM, using a key derived from your share id. The relay stores ciphertext it cannot read or forge, and it never sees the key.

Every machine

A durable copy that syncs.

One brain across your laptop, desktop and build box, with nothing ever pruned. Conflicts become copies you resolve, and daily journals merge on their own.

Yours to host

No accounts, no cloud.

Run kontexta-relay with npx or Docker on any box. Create a brain, give its share id to your other machines, and teammates can use the same relay without ever seeing each other's brain.

YOUR MACHINES
plaintext stays here
RELAY
ciphertext only
YOUR OTHER MACHINES
decrypted locally
OPERATING PRINCIPLE

kontexta turns AI agents from isolated chat sessions into a continuous operational system.

SEC. 04: INSTALLATION

Install


Choose your setup. npx, Docker and Contributors ship the dashboard and the MCP server together; MCP only runs headless from your agent's config. Relay is optional and only for syncing your brain across machines.

npx
Recommended

One command. Dashboard + MCP together. Requires Node 22 LTS.

View procedure →✓ Dashboard + MCP
Docker
Alternative

Persistent vault on a host volume.

View procedure →✓ Dashboard + MCP
MCP only
Headless

Headless MCP server from npm. No dashboard, no ports. Requires Node 22 LTS.

View procedure →MCP only
Agent Skill
No MCP

For agents whose client blocks MCP servers (managed/enterprise installs). Python bridge the agent runs from its shell.

View procedure →Shell bridge
Relay
Optional sync

Optional sync across machines. Run your own relay; it only ever stores ciphertext. Needs Node 22 LTS or Docker.

View procedure →✓ Dashboard + MCP
Contributors
Advanced

Clone, bootstrap, hack. Apache-2.0.

View procedure →✓ Dashboard + MCP
AGENT INTEGRATION: MCP INSTALLER & AUTO-JOURNALING

Connect MCP & Enable Coding Agents

Kontexta connects to your AI coding agents in seconds. Use our automated MCP installer to register kxta directly into your agent configs with zero JSON editing, and wire lightweight hooks that stream prompts, replies, and shell commands into your journal.

Claude CodeCursorGemini CLICodexCopilotClineContinueHermesAntigravityWindsurf
1-Click MCP InstallZero JSON edit
$npx kontexta connect --agent claude-code,cursor

Registers the kxta MCP server into your agent's native config file, or run npx kontexta connect to detect all agents.

Enable Journal HooksMulti-agent list
$npx kontexta hooks enable --agent claude-code,gemini

Enables agents and configures their native hooks to capture prompts, outputs, and shell actions.

Inspect & WebUIVerified status
$npx kontexta hooks status

Or connect MCP and toggle hooks with a single click in the WebUI via Configure → AGENTS or during setup.

REFERENCE SHEETS

Go deeper


Four detail sheets: the full tool inventory, the publish pipeline, an honest comparison, and per-client setup.

SEC. 05: NOTES & TOLERANCES

Questions & security


The honest answers to the things you should be asking before you point an agent at your codebase.

Where does my data live?
On your machine. By default, data lands at ~/.local/share/kontexta (Linux), ~/Library/Application Support/kontexta (macOS), or %APPDATA%\kontexta (Windows), no configuration needed. You can override this with KONTEXTA_DATA_DIR. Everything is indexed in a local SQLite database. Nothing is sent to a third-party service: no embeddings API, no telemetry. Your only outbound traffic is the optional sync to a relay you host yourself, and it is encrypted on your machine first.
Does it work offline?
Fully. The MCP server, the web UI, and the SQLite index are all local. Syncing to a relay is optional and the only operation that touches the network.
Can I use it on more than one machine? Is that private?
Yes. Connect your brain to a relay you host (npx kontexta-relay or Docker) with a share id, and your machines stay in sync. Payloads are encrypted with AES-256-GCM on your machine using a key derived from the share id, so the relay stores ciphertext it cannot read. Without a relay, nothing leaves your machine.
Which agents work with kontexta?
Anything that speaks MCP: Claude Code, Cursor, Gemini CLI, Codex, Continue, Antigravity, Claude Desktop, Aider, Cline, and OpenClaw. The admin_onboard_agent tool auto-detects and writes the right context file (CLAUDE.md, AGENTS.md, GEMINI.md, .cursor/rules/*, .continue/rules/*, .aider/kontexta.md, .cline/kontexta.md, .openclaw/kontexta.md) so agents pick up the workflow automatically.
How is Hands execution sandboxed?
Hands tools are declared per-project in kontexta.json, so agents cannot invent commands. Each invocation runs with no shell interpreter, a locked working directory, and a scrubbed environment (only PATH, HOME, USER, LANG, TZ are forwarded). Output is capped via a ring buffer; timeouts kill the entire process group. Argv parameters are validated by re2 patterns (ReDoS-proof) with an 8 KB UTF-8 cap, and the default pattern blocks leading-dash argv injection.
Can I require human approval before destructive commands run?
Yes. Mark a Hands tool as confirm-required and it returns a single-use, CSPRNG-bound token that expires in 60 seconds. The agent must call hands_confirm with that token to actually execute. The token is the auth, and it cannot be replayed.
What's the license?
Apache-2.0. Use it commercially, fork it, embed it. No telemetry callback, no rug pull.
APPENDIX A: DISTRIBUTION

Get it where you live


PEERLIST LAUNCHED
APPENDIX B: ACKNOWLEDGEMENTS

Thank You Supporters!


We launched Kontexta on Peerlist and received incredible support from the community. A massive thank you to everyone who backed our launchpad debut!

PLATE B: PEERLIST LAUNCHKontexta Peerlist Supporters
APPENDIX C: MORE LAUNCHES

Other Launches


Kontexta is live across the builder community - check out our other launchpads and platforms.

Fazier badgekontexta - Save your tokens. Remember Forever | Product Hunt
kontexta (kon-TEX-tah) · by safiyu · REV 7.2.2 · Apache-2.0