Skip to content
bucker

See also privacy notice, terms of service and the trust page.

Data processing addendum

This is the outline of the data processing addendum Bucker would sign with a customer, with each clause filled in only as far as the repository can support it. Where a clause depends on a commitment that does not exist yet — a signed sub-processor agreement, a breach notification window — it says so. The trust page and the SOC 2 readiness note say the same things at greater length, and no statement here should be read as stronger than the one there.

1. Parties and roles

  • Controller: you, the customer organization.
  • Processor: Atilan LLC, a Wyoming limited liability company, Sheridan, Wyoming, which owns and operates Bucker. Its registered address is 32 N Gould St., Suite D, Sheridan, WY 82801, USA. This outline is not a signed agreement and has not been reviewed by counsel.

2. Subject matter and duration

Processing of the personal data contained in error telemetry, session replay, logs, traces, AI-application telemetry, account data and the audit trail, for as long as the customer holds an account, and thereafter for the retention windows in §8.

3. Nature and purpose

Receiving, grouping and storing error events; projecting issues for a coding agent; running proposed fixes in a sandbox; recording human approval decisions; alerting; and the compliance functions in §7. Root-cause analysis, the issue summary and fix authoring send scrubbed issue content to a model provider (§6).

4. Categories of data subjects

  • The customer's staff and contractors who hold accounts.
  • End users of the customer's applications, whose errors, sessions and console messages are captured.
  • AI agents acting as principals under a human sponsor, to the extent their identifiers are personal data.

5. Categories of personal data

  • Account data: email address, display name, password hash, encrypted TOTP secret, directory identity where SSO or SCIM is configured.
  • Identifiers in telemetry: user ids the SDK attaches, email and IP addresses inside exception messages and request metadata, release and environment context.
  • Session replay: interaction events as described in session replay and privacy, including console messages captured verbatim up to 200 characters.
  • Audit rows: actor, delegation chain and client IP address for every action.

No special categories of data are sought. Card numbers and credentials are scrubbed before storage; their presence in telemetry is a customer-side defect, not a category processed on purpose.

6. Processor obligations

  • Instructions. Personal data is processed to provide the service as documented and for nothing else. There is no analytics, advertising or training use.
  • Confidentiality. There is no security-awareness training programme and no background check process; both are listed as not started in the readiness note.
  • Security. The technical measures in Annex II. No certification, audit, penetration test or incident-response plan exists.
  • Sub-processors. Annex III. None is contracted.
  • Model providers. Scrubbed issue content is sent to the model provider configured for the deployment, or to the one whose key the customer supplies. No zero-data-retention agreement exists with any provider.
  • Data-subject requests. Assistance is by mechanism (§7), not by a support process.
  • Breach notification. No commitment or timeline exists. A written incident-response plan is listed as not started.
  • Audit. No audit report is available. Every control claimed on the trust page names its file and its test, and the repository is available to a reviewer under agreement.

7. Data-subject requests

Access, portability and erasure are organization-scoped endpoints behind an administrator check (domain/src/modules/compliance/dsr.ts). The subject surface is derived from the live foreign-key graph, and a test asserts completeness against that graph. Erasure nulls nullable references, deletes leaf rows, pseudonymizes the subject row and points dependent rows at a pseudonymized tombstone; audit rows are retained, reduced to a principal id and a SHA-256 digest.

8. Deletion and return

Event data: a hot window (default 30 days) settable from 7 days up to the plan's retention window, floored at 30 and capped at 90 — 30 on Free, 90 on every paid plan. Event partitions are per day and shared by every project on the deployment, so a day is archived to cold storage and its table removed by DROP TABLE only once the longest plan-permitted hot window among all projects has passed — 30 days (the default) while any project is on the default, more if any project's plan permits and it has set a longer window, as short as 7 days only when every project has set a shorter window, and at most 90; one project's shorter window does not remove its own events sooner while another project's window is longer. Then cold expiry once the event's day is older than the cold window — a total age counted from the event's own day, not a further period after the hot window (default 90 days, so about 90 days in total, not 120) — settable to 0 (no archiving) or otherwise from the hot window up to the plan's retention window, floored at 90 — at most 365 days in total, on Enterprise; a stored window above the plan's limit is clamped when retention is enforced; hourly, enforced by a scheduler. Audit logs, spans, log records and replay segments: capped per plan at 30, 90, 180 or 365 days, deletion implemented and tested, swept hourly by default — unless a deployment sets COMPLIANCE_RETENTION_ENABLED=false or the runtime blob-ordering probe refuses a pass, in which case it runs only when an administrator invokes it. Portability export is the return mechanism; there is no bulk export beyond it.

9. International transfers

Processing takes place in one region, which Bucker calls us: compute runs in Dallas, United States (Fly.io dfw). The database's region has not been confirmed and the object store is a global bucket not pinned to one country; the privacy notice ("Where it is stored") and Annex III say so store by store. No EU region is in operation, and no transfer mechanism has been adopted. The intended mechanism is the EU standard contractual clauses (Module Two, and Module Three where the customer is a processor) with the UK addendum, rather than the EU-US Data Privacy Framework, under which Bucker is not self-certified (owner decision, 2026-10-17). They exist only in an unreviewed draft and are not in force.

10. Liability, governing law, term

Governing law: this addendum is governed by the laws of the State of Wyoming, USA (owner decision, 2026-10-17; not reviewed by counsel). Standard contractual clauses, if any are ever adopted, carry their own governing law and forum under their clauses 17 and 18, and this sentence does not override them.

Liability, forum and term: not drafted. Intentionally empty; nothing should be inferred from the absence.

Annex I — Details of processing

See §§2–5. The trust page lists the unauthenticated endpoints an outsider can check without an account.

Annex II — Technical and organizational measures

Each measure names the file that implements it; the trust page names the test.

Measure Mechanism
Encrypted transport to the database, enforced db/src/db-tls.ts refuses a plaintext connection to a remote host; certificates held as text from encrypted secrets, never on disk
Scrubbing before storage and again before a model shared/src/scrub/; stored tier on every ingest path, agent tier in essence, RCA and fix authoring
Human-only approval and merge domain/src/modules/approvals/requests.ts, shared/src/scopes.ts, and no merge method in domain/src/modules/scm/provider.ts
Relationship-based authorization domain/src/modules/authz/check.ts; a token's scopes are a ceiling, never the decision
One audit write path domain/src/modules/authz/audit.ts; the Human Accountability Ledger reads it back per person
Sandbox disposal on every path workers/src/remediation/execute-remediation-run.ts; default-deny egress, no production secrets, disposal in finally
Kill switches ahead of authorization domain/src/modules/authz/agent-policy.ts
Secrets encrypted in the repository, guard by hash scripts/secrets.sh, .githooks/pre-commit
Outbound request guard shared/src/egress.ts refuses private, loopback and link-local destinations
Transactional offboarding domain/src/modules/agents/identity.ts; suspending a sponsor revokes every agent it sponsors

Not present, and stated: storage-layer encryption at rest (a property of the hosting services, not documented), a written incident-response plan, periodic access review, a tested backup restore, security training, background checks.

Annex III — Sub-processors

None contracted. No data-processing agreement exists with any provider. For transparency only, the sub-processors the deployment uses or could use are listed below, generated from the one file the trust page and the privacy notice also read, so the three cannot disagree. The regions and stores they sit in are in the privacy notice under "Where it is stored" and in the trust manifest.

From the trust manifest, version 0.2-draft, as of 2026-10-28.

Sub-processor Purpose Data Location Status Contract
Fly.io, Inc. Compute for the API, the workers and the MCP server Customer content and personal data in flight and in memory; application logs United States (Dallas, region dfw) In use No agreement signed
DigitalOcean, LLC Managed PostgreSQL (the production and development databases) All stored customer content and account data Region not yet confirmed In use No agreement signed
Tigris Data, Inc. (S3-compatible object storage, provisioned through Fly.io) Object storage for event envelopes, replay recordings and cold archives, in the bucket bucker-io Envelopes, replay segments and archived events Global bucket (not pinned to one country) In use No agreement signed
Vercel Inc. Hosting for the dashboard, the marketing site and the public status site Dashboard traffic in flight; customer status-page content; no database Global edge network In use No agreement signed
Cloudflare, Inc. DNS for bucker.io (DNS-only records; no traffic is proxied) None. DNS queries for Bucker hostnames only Global In use No agreement signed
Anthropic, PBC (default model provider) Fix authoring, and root-cause analysis started automatically by an alert rule or the automatic fix sweep, on Bucker's key; on an organization's own stored key, also root-cause analysis requested from the dashboard or the API and the issue summary. Only when the organization's AI features are enabled and a hosted provider is configured. On Bucker's key, root-cause analysis requested from the dashboard or the API and the issue summary stay on the deterministic, rule-based provider Scrubbed issue content; repository source in fix-authoring prompts (not scrubbed, by design) United States Only when AI features are enabled No agreement signed
OpenAI, L.L.C. (default embeddings provider; this row names whichever OpenAI-compatible vendor is configured) Similar-issue and grouping suggestions: one embedding vector per issue, only when the organization's AI features are enabled and a hosted embeddings provider is configured Each issue's exception type, exception message, title and culprit, after the storage-tier scrub. No repository source and no event payloads United States Built, not connected on this deployment No agreement signed
Stripe, Inc. Card payment and billing portal for paid plans Billing contact and organization identifier. Card details are entered on pages Stripe hosts and never reach Bucker United States Built, not connected on this deployment No agreement signed
Modal Labs, Inc. (sandbox provider) Isolated execution of proposed fixes Repository source and test output for a proposed fix United States Built, not connected on this deployment No agreement signed

Changes to this addendum

Every change to this addendum is a dated, reviewable revision, and the history is published alongside it. Material changes will be announced before they take effect.