A local-first, zero-dependency context engine (~10.46MB) that bridges multi-gigabyte codebases to LLM reasoning windows via the Model Context Protocol (MCP).
Binaries are published automatically via GitHub Actions on every tagged release. View all releases →
5 native tools exposed over JSON-RPC 2.0 stdio — zero configuration required
*No raw files leave your machine. Only stripped structures, signatures, and sanitised text are emitted to reasoning engines.
Flat directory parsing drops full file content into LLM windows, wasting valuable tokens on repetitive boilerplates.
Agents lack context of sibling projects or packages, making signature updates extremely error-prone.
Unprocessed logs, database connections, and environment files leak directly to commercial cloud providers.
Without monitoring, agents repeatedly run identical compiler routines and cycle through file reversions indefinitely.
Filters project assets dynamically, keeping structure and signature definitions while stripping token bloat.
Maintains child-parent dependency trees with instant notification warnings when upstream APIs modify signatures.
Intercepts database queries and outputs to scrub passwords, raw keys, and proprietary tokens before payload delivery.
Monitors system changes and blocks recurring file modifications. Instantly alerts user if loop count crosses threshold.
Provides stdin/stdout communication mapping perfectly to the standard Model Context Protocol. Exposes local context directly to agent plugins.
Understands project structure at syntax level. Maps classes, structures, and declarations into SQLite index for sub-millisecond retrieval.
Establish topological linkages. Track parent-child workspace relationships and trace cascades when signature hashes change.
Auto-detection engine that intercepts outputs and cleans credentials, authentication endpoints, and SSH keys in real-time.
Self-healing safety watch loops. Flags repetitive compile or command loops to prevent token waste and infinity execution states.
Automated database compression and log rotation. Evicts dead process locks and runs vacuum routines on database files.
Neuron Core is committed to open source under AGPLv3. For proprietary environments, closed-source products, or custom integrations, we offer commercial licenses.
For developers, educators, and open-source contributors. Any derivative works must remain fully open-source under copyleft terms.
For small teams and startups (up to 25 developer seats) embedding context tools into private, proprietary workflows.
For large organizations, compliance-driven teams, and commercial vendors looking to embed Neuron Core as an OEM component in proprietary SaaS/IDE platforms.