You own the outcome.
FDEOps owns the context.
What if your AI coding agent started each session knowing who the client is, what was decided, and what's safe to ship next? FDEOps makes that possible.
Memory + methodology + skills, in one kit. Open source.
Claude · Cursor · Copilot · Codex · Gemini · Local LLMs
Three things. One @fde.
Not an app. The system of record for embeds.
Talk in plain English.
After meetings, before sponsor syncs, when scope is disputed, say it to @fde. No command cheat sheet to memorize.
See the loopOne skill that routes.
Land to close, plus overlays when the work is regulated. The agent picks the method; you confirm the judgment.
See methodsOne .fde/ per client.
Plain markdown your agent reads first and writes as you work. Private to your machine. Never uploaded.
See the fieldbookDay 1, it names what the brief left out.
Same @fde. Two weeks apart. The difference is the fieldbook.
@fde I'm on site. First meeting tomorrow. Brief: fix payments API.matched · land · validation gateLanding at Garvey. Brief says payments API. No budget owner named, no scope boundary defined. Two things before tomorrow:Who signs off on done: CTO or someone above?Has this been attempted before? Rebuilds always have a ghost in git.Your first question in that room decides whether this is a 2-week fix or a 2-month project.drafting brief.md, stakeholders.md, trust-profile.md
@fde what's the priority today?loaded context.md · 13 sessions · integrity scanIngest retry on staging, half-done from last session.CTO demo Friday. Status hasn't gone out since Tuesday, that's an amber signal.Denise (finance controller) still hasn't seen a demo: if she's surprised Friday, that's a trust fire.Order: send CTO status now, finish retry, then check Kesterman (3 days quiet, stale or dead?).One thing: your success.md says 'finance signs off' but Denise has never seen it run. Worth a 10-min walkthrough before Friday?
Open the engagement in a browser.
Your engagement rendered: trust, phase, next action, and the full record in one local HTML page. Ask @fde, or open it directly. Offline. No account.

Agent plumbing (optional): full CLI reference on GitHub.
The fieldbook. One folder per customer.
Plain markdown you own. Every claim dated and sourced, so you can defend it in any room.
Private to your machine. Never uploaded. Grep it, defend it, move it to any tool.
A loop that compounds, not a tool you configure.
No prompts to engineer. No config files. The work and the record are one action.
- 01
Describe
Plain English. 'New client, payments platform, 3-week timeline.'
- 02
Route
One @fde skill picks the right method for where you are in the engagement.
- 03
Execute
Artifacts land in .fde/. You review before anything is on the record.
- 04
Compound
Next session loads the fieldbook automatically. Context never starts from zero.
Depth under @fde, when you want it.
You don't pick these. @fde routes to the right method for where you are. Overlays activate when the engagement touches AI, exec reporting, fintech, healthcare, or government.
Embed & Trust
First days: access, credibility, scope.
Activate automatically when the engagement touches them.
Three steps. About a minute.
No accounts. No backend. No API keys. Runs fully offline once installed.
- 01
Try it
Any repo · no installnpx fdeops scan
- 02
Install @fde
Claude plugin or skills add/plugin marketplace add suboss87/fdeops
npx skills add suboss87/fdeops
- 03
Bind + start
Inside the client workspacenpx fdeops resume --init my-client
After install, talk to it: @fde I just got the brief. New client, payments, 3-week timeline.
Other agents · Cursor · Codex · Copilot · Gemini CLI: npx fdeops adapters . · Local LLMs: load skills/fde/SKILL.md as system prompt.