d627632874
Docs for parent release `v2026.1106.0` (`d69631b0`), up from `v2026.1023.0` (`1b4c8038`): 515 parent commits. ## How this branch was built The tag is 914 commits behind what `nightly` has drafted, so this branch isn't cut from `nightly` HEAD. It starts at nightly commit `1952052` (parent `043f2514`), 19 commits before the tag, then adds: - the `main` hotfix (#160) - the sync tooling from `nightly` - docs for those 19 commits Later nightly drafts stay on `nightly`, so no pages had to be removed. No `nightly-draft/*` PRs are open. ## What changed - **Runner by default:** new Claude, Codex, OpenCode and Pi agents start on Paperclip Runner on supported hosts, with **Legacy runner** and **Upgrade runner**. - **Workspaces:** cleanup of finished workspaces, optional Git expiration, and restore points for unsaved work. Restore points are off by default. - **New pages:** Prompt injection prevention, Google Workspace, GitHub Copilot, Speko, Braintrust, Daytona, Ditto and Unikraft. - **Breaking and upgrade notes:** the run history limit, Sentry run failures off by default and the DSN variables, the Docker start command for custom images, `--yes` on `instructions-file:delete`, and slower migrations on large instances. - **Also covered:** project repositories as skill sources, roster bulk edit, Artifacts V1 (experimental, off by default), composer dismissal, `expectedStatusVersion`, `unexecuted_review_stage`, and plugin workspace providers. - **Changelog:** a v2026.1106.0 entry in `reference/changelog.md`. ## Older text corrected These were wrong at the tag, and most are also wrong on the live site: - which permissions let an agent read a peer's configuration or change instructions (`api/agents.md`) - `decideNowCount`, which doesn't exist; the field is `deskBadgeCount` - `file-resources` routes, which are board-only - the **Create project** dialog fields, a CEO rule the code doesn't enforce, and where a terminated agent's reports appear - `instructions-file:put` and `:delete` examples that always failed (`cli/agent.md`) - a curl example that returned `422` (`api/cases.md`) - Google Workspace **Reconnect** instructions, which claimed it reopens the service picker - the `permissionKey` table, which listed 8 of 21 keys - drafting text at the end of `api/agents.md` ## ⚠ Please check these - **Product behaviour that may be a parent bug.** The docs now describe the tag's code, which you may prefer to fix upstream: - The project **Environment** dropdown saves a choice, but runs ignore it (`experimental/environments.md`). - `environments:manage` is granted by roles but no environment route checks it. - `suggest-changes` is documented as agent-only, because only agents have a consent path. Company skill policy decides skill edits. - **Reader-visible mismatches in the parent UI.** The docs follow the code. Upstream copy still mentions a "Built-in agents tab", says Copilot "awaits qualification", and tells people to "Reconnect Google Workspace to change access". ## ⚠ Post-tag leaks removed Nothing to remove from an already-drafted page: post-tag surfaces such as 1Password, Vercel, Supabase deploy, custom gateway onboarding and Artifacts V1 as default never came onto this branch. Authors left out post-tag hunks of the next nightly draft, and the reasons are in the run notes. ## ⚠ Drift None. Zero findings at the tag, from a full parent snapshot. ## ⚠ Verification Failures The tag-accuracy gate reports one leak and nine review entries. None is a real failure: - **Leak:** `reference/cli/service.md`, `paperclipai payload and`. This is a false positive; it's prose from a real message at `cli/src/commands/service.ts:184`. - **Review, 8 new connector pages:** the gate can't confirm new pages. Reviewers confirmed each surface exists at the tag. - **Review, `local_tools`:** an example value from the tag's Runner README. Not provable from code, and kept deliberately: - `sandbox-providers.md` uses `sandbox.yaml`. The tag's README says `install.yaml`, which returned 404 when checked. The upstream fix is post-tag. - The Google Workspace row about older OAuth brokers rests on release notes and a design doc. The broker isn't in this repo. - External claims about provider behaviour (for example Braintrust, Daytona, Unikraft and Google scopes). ## ↻ Renames detected None. The rename helper wasn't run: the 19-commit delta adds no new adapter, skill, or plugin directories that it would check. ## Verification All checks read a checkout pinned to the tag. - **First round:** authors, a release-notes audit, an audit of corrections stranded on `nightly`, an audit of 321 lines the last realign dropped from `main` (about 130 checked directly in code), and six fresh verifiers. - **Second round:** seven new verifiers re-checked every changed area, tracing behaviour rather than matching labels, and fixed about 50 more defects. That volume means more may remain. - **Limits of that second round:** - Each verifier reported what it only string-matched. That includes parts of Artifacts V1, the **Upgrade runner** notice, Copilot run behaviour, and some Plugin SDK contract text. - The second-round fixes themselves weren't reviewed by a third round. - **Gates:** `docs:build`, `docs:test` and `sync:check` pass. - **Nothing was run against a live instance.** All of this comes from reading code. ## Screenshots **196 of 278 screenshots are stale and none were recaptured.** The list is in `SCREENSHOTS_PENDING.md`. Some visibly differ from the text, such as the removed **Run with provider trace** button and the old roster tabs. ## After merge 1. Keep the PR title as is: `tag-docs-release.yml` reads the version from it. 2. Confirm the `docs/v2026.1106.0` tag exists. 3. Run `node scripts/sync/realign-nightly.mjs release/v2026.1106.0 --push`, then audit for dropped lines and duplicated paragraphs, and set `nightly` state back to `branch_mode: nightly`. 4. Check that `docs.paperclip.ing` serves the release: a new page, a new claim, and the changelog's top entry. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.8 (1M context) <[email protected]> Co-authored-by: DevRel Gemini <[email protected]> Co-authored-by: Paperclip <[email protected]> Co-authored-by: Dotta <[email protected]> Co-authored-by: ClaudeCoder <[email protected]>
8.4 KiB
paperclip_version, seo_title, seo_description
| paperclip_version | seo_title | seo_description |
|---|---|---|
| v2026.1106.0 | OpenRouter Connector | Give agents model access through OpenRouter with an API key. Harness and model ID rules, assignment, a bounded test, and fixes. |
OpenRouter
An OpenRouter connection gives agents a credential that routes to many model providers through one account. It is model access, not a set of tools — there are no actions to permit.
OpenRouter is API key only. There is no subscription sign-in for this provider, and Paperclip says so on the sign-in screen.
Before you connect
- An OpenRouter account with credit or billing set up, and an API key from OpenRouter keys.
- An agent that runs Claude, Codex, OpenCode, or Hermes. Paperclip Runner uses the same rules for its Codex, OpenCode, and ACP Claude providers.
- An upstream model ID supported by OpenRouter and your selected harness. OpenCode adds the
openrouter/harness prefix; the other harnesses use the raw upstream ID.
Connect OpenRouter
- Open Connectors and select OpenRouter.
- Paste the API key. Paperclip stores it as a secret and it is not readable afterwards.
- Optionally enter Model IDs (comma separated), then choose the account audience and agent access before selecting Connect.
Setup is a single screen with no separate access step. Once the account is saved, you change which people and agents may use it from the saved connection.
Assign the credential
Select the named account in the agent's Connection selector and save. Routed OpenRouter connections require explicit selection; they cannot become the responsible user's provider default. You can select an eligible Personal account or a Company shared account.
Each run needs a responsible user who may use the credential, and the connection must be installed for that agent or company-wide. The account owner must remain an active company member. Selecting an account does not bypass those checks.
Use the model ID shown by the selected connection. If you leave the connection's model list empty, Paperclip discovers choices from OpenRouter's public catalog and offers a refresh. You can also enter a provider ID yourself. For OpenCode, anthropic/claude-sonnet-4.5 becomes openrouter/anthropic/claude-sonnet-4.5; Claude, Codex, and Hermes retain anthropic/claude-sonnet-4.5.
Older OpenRouter accounts without routing metadata retain their OpenCode-only behavior and openrouter/ model requirement. An existing responsible-user default for that older account can keep working; reconnecting does not silently convert it to a routed connection. Create a new connection when you need the expanded harness support.
Warning: Authenticating successfully does not mean a particular model is available to you. OpenRouter decides which upstream models your account can route to, based on its own availability, your credit, and any provider-specific requirements. A key that works for one model can be refused for another.
Connect from an agent's settings
Select Connect an account… in the agent's Connection selector to open model-provider setup. Connect OpenRouter, then select the returned account and model. The connection's access settings still decide who and which agents may use it.
If a task shows an authentication card, use Fix connection to replace the stored key. Only the account owner can reconnect. The existing route and sharing stay in place, and the task resumes after successful setup.
Check your usage limits
Before you hand an agent a long task, you can see how much of the provider's allowance is left. Open the saved account and find Usage, then select Check usage. For an OpenRouter key, Paperclip asks OpenRouter for the key's credit cap, how often it resets, and any daily cap on free models, when OpenRouter returns them.
Paperclip reads the limits only when you ask. Opening the account, listing connections, and starting a run never trigger a check. Each limit window shows how much is used and when it resets, and is marked Limit reached or Blocked when the provider says so. Overage shows whether extra paid usage is available. Anything the provider leaves out shows as Not reported, never as zero. After a successful check, the button becomes Refresh.
Checking is read-only. It does not move work to another account, stop runs at a limit, refresh the credential, or buy credit.
| Message | What it means |
|---|---|
| "Sign in again to check usage." | The stored credential has expired. Reconnect the account |
| "Usage access denied." | The provider refused to share usage for this credential |
| "Too many checks. Try again later." | The provider rate-limited the check itself. Your allowance is not affected |
| "Reconnect to check usage." | The account is not usable right now. Reconnect it first |
Board users can run the same check over the API with GET /api/companies/{companyId}/ai-connections/{connectionId}/usage.
How your requests appear in OpenRouter
When an agent calls OpenRouter through Paperclip, the request names Paperclip as the app that sent it, so your OpenRouter activity shows the traffic under Paperclip. Paperclip adds the HTTP-Referer, X-Title, X-OpenRouter-Title, and X-OpenRouter-Categories headers for this. They carry only Paperclip's public product details (its name, website, and app categories) — never your company, user, run, or key. If you set any of those four headers yourself, Paperclip's values replace them; your key, model, and other routing headers are left as they are.
Try it
Reply with the single word: ready
Run it on a compatible agent after noting the selected Connection and model. OpenCode uses the openrouter/ prefix; other compatible harnesses use the upstream model ID.
A reply proves an OpenRouter credential was accepted, the harness and selected model are compatible, and the checks above passed — Paperclip refuses an ineligible binding rather than falling back to another account. It does not prove which OpenRouter account was charged; Paperclip does not surface that attribution per run, so check the activity and credit balance in your OpenRouter account if you need to confirm it.
If it fails, change only one thing at a time — the model id is the most common cause.
Note: Illustrative task, not a recorded test result.
Troubleshooting and limitations
| Problem | Likely cause | Fix |
|---|---|---|
| "Select an AI connection compatible with this harness and model" | The connection route or model does not match the selected harness | Check whether this is a routed account or an older OpenCode-only connection, then select a compatible model |
| There is no subscription option | Expected. OpenRouter is API key only | Use an API key |
| "Connect an account and choose your personal default" | The agent uses an older responsible-user binding and that person has no default | Restore that default, or explicitly select a routed account in the agent settings |
| "This run needs a responsible user to select an AI connection" | The run has no responsible person to evaluate the credential checks against | Give the work an eligible responsible user. Binding a shared connection does not work around this |
| The key works but one model is refused | OpenRouter is not routing that model for your account | Choose another model, or check the model's requirements with OpenRouter |
| Runs fail once usage rises | OpenRouter credit is exhausted or a rate limit applied | Select Check usage on the saved account to see the key's credit cap, then top up or raise limits in OpenRouter |
| Status needs attention | The key was revoked or rotated | Reconnect with a current key |
Limitations: one connection is one OpenRouter account, and it grants no tool access. A model catalog is a list of choices, not proof of account entitlement or harness compatibility. OpenRouter decides which upstream models your account can reach.
Related guides
-
Prompt injection prevention — use a shared OpenRouter connection to check tool results for injected instructions.