Reads the local project
Collects evidence from the repo, project folder, SSD/HDD, saved logs, compiler output, terminal text, and changed files.
bea-x · Coding Recovery Studio
Local recovery studio for software work.
bea-x is a GUI-first assistance tool for software work in the post-ChatGPT and post-Codex era. It reads repo evidence, changed sections, compiler output, and specialist failure cases to give humans and coding agents clearer input before the next move.
GUI-first
Coding-first
Local evidence
Agent-aware
No silent edits
Not another coding agent.
bea-x is the local evidence and recovery surface around software work. It does not try to replace the coder, invent fixes, or intervene with an agent’s work. It helps the next move start from clearer input.
AI coding tools changed how fast code moves. They did not solve how teams recover, review, and understand what just happened.
What it does
Shows what moved
Surfaces changed sections and differences made by humans or coding agents, with a history shape closer to Word or Excel than a raw database table.
Forms clearer output
Sifts compiler, build, database, network, and runtime evidence into bounded case output: what failed, what layer it belongs to, and what evidence is missing.
Local flow
The first version is meant to be simple: open a project, capture evidence, form a case, and prepare guidance for the human or agent that acts next.
1. Open projectRead local repo and project evidence.
2. Capture frameRecord changed sections and work moments.
3. Import errorsUse compiler, build, terminal, and runtime output.
4. Form caseRun specialist readers such as bea-d and bea-n.
5. Guide next movePrepare a human-readable or agent-ready packet.
First scope
For coders and project leaders
bea-x is designed to assist coders, project leads, and reviewers even when no AI agent is involved. The aim is to support the coder, human or not, with clearer project input.
- Repo evidence and changed files
- Compiler and build errors
- Database, network, mobile, and runtime failure cases
- Reviewable recovery packets
History for work, not just code
The same frame-based history can later support general editing surfaces: documents, workbook-like changes, project notes, and reviewable work moments.
- Changed sections
- Local history
- Evidence references
- Human-readable review state
Model and agent connections
Use LM Studio instead of reinventing it
bea-x does not need to become another GGUF runtime. It can communicate with LM Studio-like model runners or OpenAI-compatible endpoints when AI input is useful.
Guidance, not intervention
bea-x can prepare packets for coding agents and IDE assistants. It does not sift or silently interfere with their work. It gives clearer evidence, boundaries, and next-check guidance.
What bea-x does not do at the start
bea-x does not invent results, automatically offer corrections, rewrite source code, run a hidden coding agent, or replace the user’s model runtime. The first goal is clarified output from the local repo and its evidence.
Specialist readers
bea-x is planned as the GUI-first layer above focused local specialists.
Database failures
Reads database, migration, JDBC, Postgres, MySQL, and SQLite failure evidence without live DB access.
Network and path failures
Reads path, message, DNS, TLS, Kafka, MQTT, HTTP, Bluetooth, and device failure evidence.
More local readers
Mobile build evidence, compute/operator evidence, and work-history capture can join the same local recovery surface.
Funding and pilots
bea-x is planned as a local-first product for the next software-work era: humans, coding agents, and project leaders sharing clearer evidence before the next move.
NSC Labs is interested in early technical pilots, funding conversations, and product feedback around local coding recovery, evidence capture, and agent-aware software work.