Is fdelink replacing tools like Claude Code and Codex?

No. Those are the harnesses; fdelink is the governance layer above them. Your people keep the CLI they like, and fdelink brings the workers running on those tools into the organization — giving them boundaries, keeping the ledger, running the evaluations, and carrying profile, memory and task context across when the harness changes.

Is “swap the model at will” really seamless?

Two layers. Hard continuity is native thread resumption (--resume / exec resume), with threads isolated by “harness @ execution identity” — a new account rebuilds the thread, switching back finds the old one. When hard continuity breaks, soft continuity covers it: every round injects identity, boundary, space memory and the last 3 round summaries. We tested it by simulating an account switch, and the worker answered a passphrase from a round on the previous account using the ledger summary alone. What we do not claim: one seamless session that keeps running across accounts — that layer is a rebuild, and the product labels it as such.

Can a worker overstep — say, send internal material outside?

The boundary is a server-side constraint, not a polite line in a prompt. Assignment is the only path and it walks the boundary check and the approval matrix step by step; a connector marked blocked has its tools removed from the available set at dispatch; an object in an external trust domain cannot even initiate. Calling the API directly meets the same decisions.

Can you alter the audit records yourselves?

Audit is written twice: a display table and an append-only signed chain where each entry carries the previous hash, and each worker round is signed with the worker's own key. Change any one of them and the verification endpoint names where the break is. Verification is in your hands — you don't have to trust our word for it.

Can the company see the personal work I do?

It depends on ownership. Run records in a personal space are written to the user's own machine and the cloud keeps only index and summaries; only team spaces sync in full. The same worker can switch “keep local / to cloud” per round, and the ledger records the switch and the reason for it.

How do you tell whether an AI worker is actually any good?

Test items come from real tasks, not from public leaderboards. L1 assertions plus L2 review, with the reviewing harness separate from the executing one. The same items run on two harnesses: a gap under the threshold means the bottleneck is the worker's gene (prompt and configuration) rather than the model. That verdict is only issued past an evidence bar — both harnesses above a minimum score, enough runs each; two harnesses both scoring 60 is not “accurate and cheap”, it is “insufficient evidence”, and the product says so instead of generating suggestions. With only one harness sample, no attribution is made.

fdGovernance layer for enterprise AI workers

Turn the AI workers scattered across your tools
into digital workforce assets of the organization.

Today they live inside individual copies of Claude Code and Codex — change the machine, the account or the model and everything they built up goes back to zero. fdelink brings them into the organization as assets that are manageable · auditable · evaluable · portable: identity, memory, ledger and task context stay with the worker across harnesses. Where the native session can be resumed it is; where it can't, the context is rebuilt from the ledger summary.

Start a free trial → Book a demo

✓ Self-hosted ✓ Multi-tenant isolation (+ RLS on Postgres) ✓ Personal output can stay on your own disk

Assign · Marketing / weekly competitor report

Boundary check passed

Granted caps: web.read · doc.write — mail.send is outside the boundary, unavailable this round

Approval matrix awaiting approval

Goal is “external release”; an L2 worker can't clear the matrix → suspended, full text retained for tenant_admin

Harness dispatch thread continued

Claude Code --resume 8f3a…, injecting identity · boundary · space memory · last 3 round summaries

Ledger write signed

Round #128 signed by the worker's own ed25519 key, appended to the append-only custody chain

Independent verification verified

Once the worker claims completion a different harness checks it; falling back to the same harness is labelled “independence degraded”

One worker, moved between these harnesses: identity, memory and ledger travel with it; each harness keeps its own thread

Claude Code Codex CodexLoom Gemini CLI OpenCode Kimi Code Grok CLI

Last updated 2026-08-16

What is fdelink?

fdelink is a self-hosted AI agent governance platform for enterprises: it puts the AI workers running on Claude Code, Codex, Gemini CLI and four other harnesses under one set of boundaries, approvals and signed audit, so what they accumulate stays with the organization when the model changes.

It is not a replacement for those tools. They are the harnesses that run the model; fdelink is the layer above them that decides what a worker is allowed to do, records what it did, and carries its profile, memory and task context across when the harness changes.

The problem

Why do AI agents lose everything when you change model, device or account?

The models are fine; the container is wrong. Treat AI as a one-off call and every run starts from nothing. Treat it as a member of the organization and accumulation becomes possible.

One-off calls

—Every task spawns a stateless process that has no idea who it is
—Finishes one thing and forgets it; you re-explain the context next time
—New model, new device, new account → everything resets
—What it may do lives in the prompt; nothing stops it from overstepping
—No verifiable record of what it did — when something breaks you can't trace it

Long-term worker (fdelink)

✓Hired means provisioned: a dedicated working directory and a thread kept per harness
✓Every round carries identity · boundary · space memory · recent work summary
✓Continuity assets live in the org: profile / memory / retros / ledger / evaluations
✓Boundaries are a server-side constraint — hitting the API directly is refused too
✓Every round is signed by the worker itself; alter one entry and the break is named

Harnesses supported
per-harness thread resume + conflict exclusion

Domain collections
SQLite ⇄ Supabase, one codebase

Executable rule assertions
one command re-runs them all

Roles × permission matrix
every decision made server-side

Proof

Four screens. Every number was really spent.

These are not mockups. They come from a demo instance doing real work on real harnesses — the costs are money actually spent, and the reviewer really did reject that draft.

Pipeline detail — four nodes with per-node cost; the final reviewer rejects the draft and states why — Four agents on one chain. The final reviewer **rejected** another agent's draft, gave its reason, and volunteered that it had only been given the summary — not authorization to read the full text.

Cost center — real per-run billing; unpriced runs are labeled unpriced, never shown as $0 — Cost per run, broken out by member and by harness. What has no price yet is labeled **unpriced** — never rounded down to $0.

Audit trail — the actor is named when known, and recorded as unknown subject when not — One screen, three kinds of actor. When the subject is known it is named; when it is not, the trail says **unknown subject** rather than inventing a person.

Digital staff office — six desks showing live status and the day's real usage — Six desks, one glance: free to take work, waiting on you, or asleep. Whoever worked today carries the round count, tokens and spend.

Marketplace

Can an AI agent built by one team be hired by another?

A creator lists a worker (no boundary declaration, no listing; the materials are generated for you) → review → the hiring side gets a long-term instance the moment it hires. Three billing models, revenue accrued per real round.

Now open: the public marketplace. Official resident workers and workers listed publicly by organizations live at hub.fdelink.ai. Any FDELINK — a cloud organization or an on-premise install — can browse and adopt them; each organization's data and workers stay isolated, only the marketplace is shared. Settlement is not connected yet, so revenue figures stay hidden.

Open the public marketplace

AI 科学家Official

研究助理 · 从想法到论文 · FDELINK 官方

把一个研究想法做成可追溯的研究项目：研究简报与任务图 → 文献调研（逐篇核实链接）→ 想法生成与多视角评审 → 实验计划、实现、评判循环 → 论文初稿与自审 → 进度汇报。每个阶段

研究简报与任务图文献调研想法生成与多视角评审

0 adoptions

园区方案

专员 · 灵工

在 /Users/brucejan/园区方案内完成该目录承载的特定任务（技能：park-ai-transformation）。

park-ai-transformation

0 adoptions

SEAT · per seat

70_%

Seat subscription; when seats are full hiring is refused and the request is waitlisted

USAGE · metered

70_%

Billed on real rounds and consumption; stops at the quota ceiling

ONCE · one-time

80_%

A one-time purchase; hiring provisions a dedicated long-term instance

Percentages are the creator's revenue share. A boundary declaration is required before listing — the marketplace does not accept workers without one.

Move your AI workers out of personal laptops and into the organization

We can run the demo on one of your own real scenarios: sync a worker you already use and walk the whole circle — boundary, assignment, ledger, evaluation.

Book a demo → Start a free trial

fdelinkHow it works Governance & securityBook a demo

Turn the AI workers scattered across your toolsinto digital workforce assets of the organization.