paperclipai/paperclip-docs

0 Fork 0

mirror of https://github.com/paperclipai/paperclip-docs.git synced 2026-11-07 06:16:07 -10:00

Files

T

d627632874 Release docs for paperclip v2026.1106.0 (#172 )

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]>

2026-11-07 13:02:42 +01:00

7.3 KiB

Raw Blame History

paperclip_version, seo_title, seo_description

paperclip_version	seo_title	seo_description
v2026.1106.0	Braintrust Connector	Let agents explore Braintrust traces, datasets, and evaluations. Pick the US or EU region, sign in or paste a key, and gate dataset edits and automations.

Braintrust

You can let agents work with your AI evaluation data in Braintrust: query traces with SQL, inspect experiments, manage datasets and scorers, and run evaluations. It is useful when a task needs to check how a model or prompt is actually behaving, not just how it was meant to behave.

Braintrust appears in the catalog on every instance; no experimental setting is needed.

The connection uses Braintrust's hosted MCP server. Every method is read and write: agents can change data and start billable work as well as read.

Before you connect

Know your organization's region. Braintrust hosts data in the US or the EU, and Paperclip offers a separate pair of methods for each. Pick the region your organization's data lives in. Switching regions will not fix a credential error.
Pick how to sign in. Browser sign-in uses your Braintrust account and needs nothing set up first. For an API key, create one in Braintrust under Settings → API keys. A separate key for Paperclip is easiest to remove later.
Decide which agents should see the data. Traces and datasets often hold prompts, model outputs, and customer content.

Access always follows the Braintrust account or key you connect. Paperclip cannot widen or narrow those permissions, and it does not add a project boundary of its own: a project named in an agent's request is not a limit.

Running a self-hosted Braintrust? Use Custom MCP servers with your deployment's MCP address instead.

Connect Braintrust

Open Connectors and select Braintrust.
Read the access line above the main button. It says who the connection signs in as and which agents can use it. Select Change to narrow the agents or to pick a different method.

Choose the method that matches your region and how you want to sign in:

Method	Region	How you sign in
Sign in with Braintrust (US)	US	Browser sign-in (the default)
Use a Braintrust API key (US)	US	API key
Sign in with Braintrust (EU)	EU	Browser sign-in
Use a Braintrust API key (EU)	EU	API key

Before you sign in, Paperclip shows a summary of what the connection can do, including the costs, permanent deletes, and automation risks covered below. Read it.
Complete browser sign-in, or paste your API key. For browser sign-in, Paperclip registers its client with Braintrust automatically, so there is nothing to set up in a developer console.
Finish setup and open Permissions to review the action list.

Paperclip stores the key or sign-in as a secret and sends it to Braintrust on each call. Agents never see it.

Choose access

Paperclip classifies every Braintrust action by name, using a list it reviewed against Braintrust's tool reference:

Reads: SQL queries, schema and experiment summaries, object and trace lookups, monitoring charts and views, documentation search, and loading a Braintrust skill.
Destructive: edit_dataset_rows, because one action can insert, replace, or permanently delete rows, and set_automation_status, because turning an automation on can apply data retention or send data elsewhere.
Writes: everything else, including evaluations, evaluator and scorer tests, dataset additions, dashboards, alerts, and Topics processing. Any action Paperclip has not reviewed is also treated as a write.

A Braintrust annotation can raise an action's risk but never lowers a reviewed one.

The classification is a label, not a lock. Discovered actions still start as Allowed, like on any new connection. Before agents run unattended, open Permissions and set the two destructive actions to Ask first or Off. Consider doing the same for evaluations and Topics processing, which can use Braintrust or model credits. See Set action permissions.

Two more things are worth knowing:

Changed actions are held for review. After the first discovery, any action Braintrust adds or changes is held back until you approve it, instead of switching on. See What happens to a brand-new action.
Your choices survive reconnecting. Actions you set to Ask first or Off stay that way when you reconnect or refresh the action list.

Try it

Start with a read you can check in Braintrust yourself:

Using Braintrust, list the five most recently updated projects or experiments this connection can see. Do not run evaluations or edit any datasets.

Compare the answer with Braintrust's own view of the same organization.

Note: Illustrative task, not a recorded test result.

Disconnect

Removing the connection in Paperclip clears its stored credentials and stops agents using it. If you used an API key, also deactivate the key in Braintrust so copies elsewhere stop working. Braintrust does not publish a way for Paperclip to revoke a browser sign-in on its side, so removing the connection does not cancel that sign-in in Braintrust. See Reauthorize, revoke, or disconnect.

Troubleshooting and limitations

Problem	Likely cause	Fix
The key is rejected	It expired, was deactivated, or was copied incompletely	Create a new key under Settings → API keys and replace it — see Replace an API key
Sign-in fails or the connection needs attention	Access was withdrawn or expired	Reconnect with the same region and method
An action fails with a permission error	The Braintrust account or key lacks access	Connect an account or key with the Braintrust access you intend to allow
An agent is waiting on an action	It is set to Ask first	Approve or decline it — see Answer a connector review request
A new action is not available	Changed actions are held for review	Select Refresh actions on Permissions and review the new entries

Limitations: one region per connection. Paperclip has no project filter for Braintrust. This connector does not send Paperclip's own traces to Braintrust.

Enterpret and PostHog — other data and analytics connectors.
How connector access works
Set action permissions
Verify a connector and fix a broken one
Braintrust MCP documentation

Sources

Braintrust definition — methods, regions, endpoints, key placement, and warnings.
Braintrust risk rules and connector policy registry — reviewed reads, destructive actions, write default, and held-back changed actions.

7.3 KiB Raw Blame History