TrueGrit AI · truegrit.dev
Agents that lend people true grit.
We build domain-expert AI agents that enable humans — bringing grit and confidence to the people facing the hardest challenges of their craft. The agent does the relentless part; you stay the expert in charge. Two arenas so far: the wet lab and the courtroom.
Humans stay in charge
Our agents enable people; they never replace them. The scientist signs the experiment and the attorney signs the brief. Anything that leaves your machine passes through you first.
Grit is verifiable
Confidence comes from provenance, not bravado. Every protocol, citation, and draft traces back to a real source — nothing important is ever pulled from model memory alone.
Local-first
Your files stay on your disk and your work runs on your own machine with your own AI subscriptions. No lock-in, no data leaving home without your say-so.
Product — At the bench
Labee
The wet-lab protocol research agent.
Bench research runs on protocols — and protocols deserve better than a binder. Labee is a Claude-powered research workspace that captures your wet-lab protocols as a living skills catalog, then puts an agent beside you to plan, run, and reason through the work.
Protocol sheet · Labee workspaceRev. current
-
Step 01
Skills catalog of reusable protocols
Capture research protocols once as executable skills, then run, adapt, and share them across every experiment.
-
Step 02
Chat workspace that speaks protocol
A Claude-backed agent to design experiments, troubleshoot runs, and research methods without leaving your workspace.
-
Step 03
Interactive canvas
Visualize results and work through analysis in real time, side by side with the agent.
-
Step 04
File deck for your data
Per-user file management keeps raw data, notes, and results organized and attached to the work.
-
Step 05
Desktop and web, your infrastructure
Run it as a desktop app or deploy it as a server — local SQLite or your own PostgreSQL.
Product — At the bar
AttyMate
A local-first AI law firm for California civil litigation.
AttyMate puts a supervised team of ten legal agents behind a single front door: intake, facts and evidence, research, drafting, QA, docket checks, and calendaring — coordinated on a kanban board while you keep every decision that matters.
It is built the way a careful firm works: source-bound (no authority ever comes from model memory — citations are researched and independently re-verified), matter-sealed (no cross-matter carryover), and gated so that filing, serving, sending, or spending always waits for the attorney.
AttyMate · Capability recordSup. atty. review req'd
-
§ 1
End-to-end litigation drafting
Motions, separate statements, declarations, proposed orders, and proof of service — drafted section by section into Word.
-
§ 2
Subpoena motion-to-compel workflow
A first-class autonomous workflow from objection table and meet-and-confer record through the full MTC package.
-
§ 3
Source-bound research & citation QA
Lexis-backed research, independently re-verified by a QA agent. Memory-derived authorities are structurally prohibited.
-
§ 4
Pleading intake with provenance
Local OCR intake of complaints and exhibits with checksums and a source-to-work-product map for every artifact.
-
§ 5
Docket checks & calendaring
Scheduled public docket monitoring and computed deadlines — proposed to you, never written without you.
-
§ 6
Matter Safety Contract
Every task carries machine-enforced scope: one writable output root, read-only sources, sealed matters, and red gates for anything external.