Project agents
→Claude Code ProjectsCodex cloud tasks
A coordinator, and workers on any model.
One conversation per goal. Each worker picks its own provider and model and runs on the project’s machine, and you can step into any of them.
SpecWithCode Free, in early access Every AI agent.
The open-source CLI SpecWithCode builds on. MIT, local first, no account.
Every AI agent.
One workspace.
Projects like Claude Code. A personal agent like Dots or Hermes. Coding agents from Anthropic, OpenAI, xAI, Google, GitHub and Cursor. SpecWithCode runs them as one project on machines you own, gives each job the model that fits, and keeps the record in git so any agent can carry on.
npm install -g specweaveWorks withOne workspace, any provider ↗
Claude CodeClaude Code ProjectsCodexGrok BuildCursorGemini CLIGitHub CopilotOpenCodeClaude CodeClaude Code ProjectsCodexGrok BuildCursorGemini CLIGitHub CopilotOpenCode
01 / The umbrellaMany agents.
Many agents.
One place to run them.
Every lab now ships a project agent, a personal agent and a coding agent, each in its own app and its own cloud. SpecWithCode gives you all three kinds in one workspace on your machines and runs every lab’s coding agent side by side, so each job gets the model that fits it, not the one from the app you happen to have open.
In the app
Personal agents
→OpenAI DotsxAI Grok BotMeta Muse
Personal projects that keep working.
A project with its own folder, memory and plan, scheduled routines, a Telegram channel to its coordinator through your own bot, and an opt-in browser profile with your saved logins.
Runs while your machine is awake, not in a vendor cloud. Webhook-started tasks need remote access turned on.
Self-hosted agents
→Hermes AgentOpenClaw
Local first, with tools you grant.
Agents run on your machines. Tools come from local MCP servers you approve one tool at a time, and memory is plain files in the project.
Telegram works with your own bot; WhatsApp and Slack do not yet.
Coding agents
→Claude CodeCodexGrok BuildCursorCopilot
Every lab’s official agent.
SpecWithCode drives each provider’s own agent, so its sign-in, approvals and limits apply. Switch a thread to another provider mid-conversation.
Usage limits
→One usage page per account
Every account’s limits on one screen.
When a provider runs out, a thread offers another provider, waits for the reset or switches on its own; you set the rule per project. With the optional Switch companion, a capacity policy drafts new work on a fresh account.
The record
→Memory locked in one vendor
Spec, evidence and handoff in git.
The open-source SpecWeave CLI keeps it, and any tool continues with specweave pickup.
Other products as described on their public pages and in launch coverage, October 2026; they change quickly. Status is SpecWithCode’s own: In the app means it is in the current build.
02 / SpecWithCode projectsOne goal.
One goal.
Many agents.
A project is a long-running conversation with a coordinator. Give it a goal, add tasks as they come up, and step into any worker to steer it.
One conversation per goal.
Tell the coordinator what you want. It splits the work, starts or reuses workers, checks what they bring back and asks you only when a decision is yours.
Each role picks its own model.
Run the coordinator on Claude Code and a worker on Codex, Grok Build or Cursor. SpecWithCode drives each provider’s official agent, so its own sign-in, approvals and limits apply, and you can switch a thread to another provider mid-conversation.
Said once, known by every worker.
Each project keeps its own memory, instructions and settings. A decision you give the coordinator reaches the next worker without you repeating it.
Reported done is not verified.
An optional plan holds the spec, tasks and assignments, or reads an existing SpecWeave increment. A task is verified only when its check runs and passes.
A device folder plus several repositories.
Point a project at a folder on a machine, add approved neighbouring folders and the repository URLs it works across. Workers get their own worktrees.
Recurring work, not only code.
Schedule routines inside a project. A Personal project gets its own working folder and per-project tool grants for research, notes and everyday tasks.
03 / Local firstYour machines.
Your machines.
Not a cloud container.
Cloud project agents are simple to start and live with one vendor. SpecWithCode keeps the work where your code, skills and accounts already are, and lets you mix vendors inside one project.
| Topic | Cloud project agents | SpecWithCode |
|---|---|---|
| Where work runs | A cloud container per thread, with a bridge to your computer in some products. | On machines you own: this computer, plus any you pair over your network, SSH or Tailscale. |
| Models | One vendor’s models. | Claude Code, Codex, Grok Build, Cursor, OpenCode, Copilot and more. The coordinator and each worker choose their own. |
| Your skills and folders | Upload skills to the account or a repository. Neighbouring local folders are out of reach. | The agent runs on your machine with its provider’s local configuration, so local skills, MCP servers and folders you approve are within reach. |
| When a limit hits | Wait for the reset or change plans. | By default the thread offers to continue on another provider, carrying the recent conversation over; per project you can have it wait and resume on the same account, or switch on its own. Or hand off to another tool with SpecWeave. |
| The record | Inside the vendor’s project. | Specs, task evidence and handoffs in git, readable by any tool. |
| Laptop closed | Keeps running in the cloud. | Runs while the executing machine is awake, so a desktop or Mac mini can carry work while your laptop sleeps. |
Cloud project agents include Claude Code Projects, OpenAI Dots and xAI Grok Bot. Compared from their public documentation in October 2026; these products change quickly.
Read the SpecWithCode overview ↗04 / Switch tools. Keep your place.The work lives in your repo.
The work lives in your repo.
Tools come and go.
- Step 1 · Write it down once
One file says
what done means.Each piece of work is a single
spec.md: the problem, the scope, acceptance criteria, the approach and the tasks. No second file to keep in sync, and SpecWeave never writes state back into it.
.specweave/increments/0042-checkout-recovery/spec.md# Keep checkout resumable ## Problem A customer who leaves checkout loses their choices. ## Acceptance Criteria - [ ] AC-01: Returning within 24h restores the cart - [ ] AC-02: A paid order is never restored ## Tasks ### T-01 Save the checkout draft - AC: AC-01 | Files: src/checkout/draft.ts | Test: npm test -- draft ### T-02 Restore it on return - AC: AC-01, AC-02 | Files: src/checkout/restore.ts | Test: npm test -- restore
- Step 2 · Work one task at a time
Each task arrives
with its criteria.The CLI gives the agent the next task with the text of the acceptance criteria it covers, its files and its test, so it reads a few lines instead of the whole spec again. A criterion is met when every task that covers it is done.
terminal · Claude Code$ specweave task next T-02 Restore it on return AC-01: Returning within 24h restores the cart AC-02: A paid order is never restored Files: src/checkout/restore.ts | Test: npm test -- restore $ specweave task claim T-02 Claimed T-02 as claude@laptop $ specweave task done T-02 --run "npm test -- restore" Done T-02 (2/3)
- Step 3 · Hit a limit
Out of tokens?
Hand it off.Tell your agent to hand off, or run
specweave handoff. It releases your task claims, records where you stopped, and pushes your branch and a snapshot of your uncommitted edits to git. Nothing to copy, nothing to paste. Withspecweave auto-handoff on, Claude Code and Codex flag it at 95% of the limit and hand off on their own when a turn hits it.
terminal · Claude Code$ specweave handoff Handed off 0042 (released T-03, pushed feature/checkout-recovery, pushed your uncommitted edits). To continue in any tool, machine or account, say "pick up" there (or run `specweave pickup`). Details: .specweave/increments/0042-…/handoff.md # what the next session reads # spec.md criteria and tasks # ledger.jsonl 2 of 3 done, T-03 free # handoff.md where you stopped # wip/… your uncommitted edits
- Step 4 · Continue anywhere
Another tool.
Same place.In Codex, Grok, Cursor or a second Claude subscription,
specweave pickupfetches the handoff, applies your edits, and prints everything the new session needs in one read: the increment, the next task and its criteria, notes and project memory.
terminal · Codex$ specweave pickup SpecWeave pickup · you are codex@laptop Picked up the handoff from claude@laptop. Increment 0042 "Keep checkout resumable" (active) Next: T-03 Drop drafts after payment AC-02: A paid order is never restored Files: src/checkout/expiry.ts | Test: npm test -- expiry claim: specweave task claim T-03 0042 Notes: - claude@laptop 5m ago: restore works; expiry not started Memory (.specweave/memory/MEMORY.md): Drafts live in Redis with a 24h TTL; never in Postgres. Spec: .specweave/increments/0042-checkout-recovery/spec.md
05 / The stackOne workspace.
One workspace.
One record.
SpecWithCode is where you work. SpecWeave keeps what survives the session in git. vskill brings tested expertise to every agent.
SpecWithCode
Projects with a coordinator and workers, each on the provider and model you pick, running on your own machines with approvals you can see.
See SpecWithCode ↗SpecWeave CLI
The spec, task claims, test evidence and handoff live in your repository, so any agent or account can continue. Open source.
Start the open-source workflow ↗vskill + Skill Studio
Inspect and install skills into every agent. Evaluate their behavior against a baseline before your team relies on them.
Explore skills and evaluations ↗06 / Parallel workOne thread.
One thread.
One increment.
A SpecWithCode worker or a Claude Code Projects thread works on one branch and opens one pull request. So does an increment. SpecWeave maps the two, and keeps everything that matters in git, where every tool and every account can read it.
How threads map to increments ↗In a Claude Code ProjectIn SpecWeave 3.0
A projectAGENTS.md and the project goal
Project memory.specweave/memory/, committed with the code
A threadOne increment: one branch, one pull request
The thread checklistThe Tasks section of that spec.md
A note to another threadA note in the other increment's ledger
07 / Fit your teamYour tracker.
Your tracker.
On your terms.
Keep GitHub, Jira or Azure DevOps where your team plans. In 3.0, changing an increment's status no longer creates or closes issues. SpecWeave touches a tracker only when you push to it.
Explore optional integrations ↗GitHubFirst-class: one issue per increment, updated on push↗JiraOpt-in: created and updated on push, closed on complete↗Azure DevOpsOpt-in: work items created on push, closed on complete↗
$ specweave sync push --dry-runPreview what would change. Nothing reaches a tracker until you run it without the flag.08 / In practice
Built with SpecWeave.
Products where we use the workflow to plan changes, track delivery and carry work between sessions and subscriptions.
Sports platforms
More projects and how we use SpecWeave ↗EasyChamp
League and club tools, built across Claude Code threads, two subscriptions and Codex.
Career toolsJobWeave
Job search, tailored resumes, recruiter context and interview preparation.
Developer toolsVerified Skills
Skill discovery, source inspection and local evaluations with Skill Studio.
The companion project
Explore Verified Skills ↗Better skills. Less baggage.
Find focused expertise, inspect its source and evaluate whether it helps your workflow.