DenlyLive Chat
Open a project-scoped coding session and follow tools and messages in real time.
Free plan · no card required
Your own coding agents, on your own machines, on the subscription you already pay for.
Describe a task. Your agent – Claude Code, Codex, Cursor, Grok, or Antigravity – plans it, you approve, it builds and tests on your own machine, opens a PR. You review and merge, all from one dashboard, even your phone.
No YAML, no resold tokens. One CLI on every supported platform.
Your machineThe agent runs on your own machine. Source never touches Denly's servers.
local
Your subscriptionBring your own agent & plan. We never mark up tokens.
yours
Your gatesPlan → you approve → build → you review the PR. Never merges alone.
your move
Honest scope
Denly CLI · macOS, Windows & Linux
GitHub & GitLab repos
Optional end-to-end encryption
token billing
The trade every agent forces
Today you pick your poison.
Coding agents come in two shapes, and both ask you to give something up.
Someone else's cloud
Cursor-in-the-cloud, Devin & friends
Convenient, always-on – but you ship your source to their servers and buy inference back at a markup. Your code and your bill are both in someone else's hands.
Local, but hands-on
A CLI agent you babysit
Your code stays home, but you're parked in a terminal watching one task at a time, answering prompts, re-running checks. It doesn't scale past a single file.
Denly keeps the local privacy and gives you the fleet – async, supervised, on your own hardware.
The task lifecycle
Six steps. Two of them are yours.
The agent does the work; you hold the gates. It stops and waits for you at every decision that matters – amber means it's your move.
1
You describe
Write the task in plain English
“Add localization (EN / ES / FR).” No labels, no board setup, no ticket template. Point it at a repo and go.
2
Agent works
Your agent reads the repo and posts a plan
Approach, the files it intends to touch, the tests it will run, and the risks – written out before a single line changes. Use Claude Code, Codex, Cursor, Grok Build, or Antigravity, per project.
3
Your move
You approve the plan
Nothing is built until you say so. Approve it, or send it back with notes and the agent plans again. It can also stop mid-run to ask a question and wait for your answer.
4
Agent works
It builds & tests on your machine
Writes the code, runs the checks you configured for the repository, pushes a branch like agent/task-28 – all on your own machine, against your real toolchain.
5
Your move
You review the pull request
A normal PR with a normal diff and green checks. Read it, comment, request changes – the same review you'd give a teammate.
6
Done
Merge
Merge from the dashboard – or your phone. The agent never merges behind your back.
What you see
One board. Every task, and who it is waiting on.
Each task rides the same rail – describe it, the agent plans, you approve, it builds, it reviews, you merge. The card tells you which step it is on and whose move it is, so the ones that need a decision are obvious and everything else moves along on its own.
Amber = waiting on you
Teal = done, or the agent's
More control when you need it
One lifecycle, with the supporting system built in.
Scoped Skills & Docs
Give each task only the reusable instructions and repository context it needs.
Flows, reviews & schedules
Add custom steps, human gates, multiple reviewers, and recurring work.
Teams & assignments
Share private projects safely and route tasks to the intended teammate.
Multi-repository runners
Use project groups and one machine runner for coordinated work across repositories.
Usage, cost & insights
Track reported quota, tokens, budgets, outcomes, and runtime success.
Integrations
Create agent tasks from your tracker or chat tool – progress, comments, and status flow back automatically. All 15 providers, with a setup runbook each:
The data boundary
Your source repository is never uploaded to Denly.
Denly's servers never receive your source in readable form: code is cloned, edited, built, and tested on your runner, not uploaded to us. Optional workspace end-to-end encryption adds a second boundary: when enabled, task text, plans, conversations, and attachments leave the browser or runner only as ciphertext unreadable to Denly. Two parties remain outside that boundary: your Git host and the AI provider you chose still receive the data the work requires under your agreements with them.
Band 1Never reaches Denlynot even encrypted – it is simply never sent to us
Your source codecloned, edited, and tested locally
Your AI key & subscriptionread from your machine only
The agent sessionthe CLI runs entirely on your machine
Band 2 · optionalCiphertext when workspace E2E is enabledend-to-end encrypted under a passphrase only you hold
Task titles & descriptions7b1a·9f…c4e2
Plans, comments & agent conversationse0d3·2a…81bf
Attachmentsencrypted before upload when E2E is on
Band 3Plaintext metadata we seethe relay needs it to route and bill – never content
Task status & timingstage, progress, when it ran
Cost & token numbersthe figures, not the traffic
PR linkspointers to GitHub, not diffs
Who and what ran itassignee, runtime and model, repository name, flow configuration
We hold no copy of your passphrase.
End-to-end encryption is only real if we can't decrypt. We can't – which also means if you lose your passphrase, that encrypted history is unrecoverable. That's the cost of a key only you hold.
The honest trade-off: work runs while your machine is awake.
Because the agent runs on your machine, tasks execute only while that runner is awake and online. Queue work from the dashboard while it is offline and it runs when the runner wakes. That's the direct cost of using your subscription: a coding-agent login is bound to your machine, so “uses my subscription” and “runs in the cloud 24/7” can't both be true.
Setup, end to end
Four accurate steps, then your first task.
No config file, no YAML, no labels to create, no board to set up, no gh auth login.
1
Sign in
GitHub, Google, or an email and password. That's the whole account step.
Continue with GitHub
2
Connect a repository
Choose GitHub, GitLab, or a manual repository. GitHub App access is scoped to the repositories you pick.
storefront
added
3
Install and sign in to the CLI
Follow the platform-specific Quickstart, then run denly login on the machine that will execute work.
denly login ✓
signed in
4
Install the project
Copy and run the dashboard-provided denly install <project-id> command. It clones or adopts the repository, starts the runner, and verifies a live status.
denly install storefront
Then it clones the repo and starts the runner for you – zero config. Describe your first task and the agent picks it up. Every command in one place: the denly CLI reference.
Pricing
Start free. Upgrade when your work grows.
Your plan covers your Denly account and the workspaces you created. Denly charges for the control plane only; AI usage stays directly with its vendor.
Prices shown in USD. Your local currency and taxes are confirmed at checkout.
The plan limits below apply. Nothing you built is ever deleted, and nothing is frozen for storage: past an allowance, new uploads and document saves are refused and everything already there stays readable and keeps running. A change that reduces what your plan covers gets 60 days’ notice, never before the end of a term you have already paid for.
Free
$0forever
For trying Denly on real work.
- Unlimited projects
- 2 machines, one task at a time
- 250 MB of documents and attachments
- 1 person, 1 workspace
- Your own agent subscription or API key
Pro
For individuals$10/ month
For running a real fleet across machines.
- Unlimited projects and machines
- Tasks in parallel, not one at a time
- 5 GB, pooled across your workspaces
- 1 person, unlimited workspaces
- Your own agent subscription or API key
Teams
2 seats minimum · 14-day trial$12per seat / month
For teams coordinating multiple products.
- Everything in Pro
- 10 GB + 2 GB per seat, pooled
- Seats are people, not workspaces
- Role-based team access
AI cost is always yours – we never mark up tokens.
There is no token billing, ever. You bring your own agent subscription or API key; inference is billed to you, by the vendor, at their price. We don't touch it.
Straight answers
The questions a skeptic actually has.
Your source repository is never uploaded to Denly; it is cloned, built, and tested on your runner machine. When optional workspace E2E is enabled, task text, plans, conversations, and attachments are encrypted before upload and unreadable to us. Your Git host and chosen AI provider remain outside that boundary.
Claude Code, OpenAI Codex, Cursor, Grok, or Antigravity – on your own subscription or API key. Set it per project or as a workspace default.
Never. Inference is billed to you directly by your agent's vendor. Denly's Free plan is $0, Pro is $10/month, and Teams is $12 per seat/month with a 2-seat minimum and a 14-day trial – a seat is one person, however many of your workspaces they are in.
Work runs only while your runner machine is awake and online. Anything queued while it is offline waits and runs when the runner wakes.
Yes. macOS, Windows, and Linux use the same Denly CLI. On Windows, one PowerShell command installs it for your user with no admin rights. After
denly login, the dashboard shows the exact denly install <project-id> command to run.Exactly two: Contents and Pull requests, scoped to the repos you pick when you install the GitHub App. That's all – no org-wide access, no admin.
No. It plans, builds, tests, and opens a PR, then stops. You approve the plan, and you review and merge the PR. You hold the gates by design.
Point it at a repo and close the dashboard.
Start free with no card. No token markup, no config file.