- feat(dashboard): what needs a person, what broke, and the day ahead come first (#862)latest8Docs93688682026-10-31 07:19
Change detail
8 changes across 2 operations
- GET/v1/workspaces/{workspace}/dashboard
- info
the
response's property default value changed fromlinear-mtdtotrailing-7dfor the status - info
added the required property
///to the response with the status - info
added the required property
///to the response with the status - info
added the required property
to the response with the status - info
added the required property
to the response with the status - info
added the required property
to the response with the status
- info
- GET/v1/workspaces/{workspace}/balance
- info
added the optional property
/to the response with the status - info
added the optional property
/to the response with the status
- info
- feat(runs): reasoning levels come from the catalog; a reasoning token budget is a setting (#849)9Docs9bdc0812026-10-30 08:54
Change detail
9 changes across 7 operations
- PATCH/v1/workspaces/{workspace}/agents/{agent_id}
- info
added the new optional request property
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/
- info
added the new optional request property
- info
added the optional property
to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents/{agent_id}
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/{agent_id}/install
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/{agent_id}/restore
- info
added the optional property
to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents
- info
added the optional property
/to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents/
- info
added the optional property
/to the response with the status
- info
- fix(runs): a reply cut off at the output limit is continued, not taken as the answer; reasoning effort is a setting (#845)9Docsc175ff22026-10-29 21:05
Change detail
9 changes across 7 operations
- PATCH/v1/workspaces/{workspace}/agents/{agent_id}
- info
added the new optional request property
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/
- info
added the new optional request property
- info
added the optional property
to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents/{agent_id}
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/{agent_id}/install
- info
added the optional property
to the response with the status
- info
- POST/v1/workspaces/{workspace}/agents/{agent_id}/restore
- info
added the optional property
to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents
- info
added the optional property
/to the response with the status
- info
- GET/v1/workspaces/{workspace}/agents/
- info
added the optional property
/to the response with the status
- info
- feat: agent settings polish, per-tool workspace defaults on the Tools tab, platform connections keep their address private (#813)41Docs17ff81c2026-10-27 23:56
Change detail
5 changes across 5 operations
- GET/v1/workspaces/{workspace}/openapi-connections/{connection_id}
- breaking
response property
base_urllist-of-types was widened by adding typesnullto media typeapplication/jsonof response200
- breaking
- PATCH/v1/workspaces/{workspace}/openapi-connections/{connection_id}
- breaking
response property
base_urllist-of-types was widened by adding typesnullto media typeapplication/jsonof response200
- breaking
- GET/v1/workspaces/{workspace}/openapi-connections/
- breaking
response property
items/base_urllist-of-types was widened by adding typesnullto media typeapplication/jsonof response200
- breaking
- POST/v1/workspaces/{workspace}/openapi-connections/
- breaking
response property
base_urllist-of-types was widened by adding typesnullto media typeapplication/jsonof response201
- breaking
- GET/v1/workspaces/{workspace}/mcp-server-instances/{instance_id}/tool-policies
- info
endpoint added
- info
- feat(analytics): anonymous usage analytics with one switch (ADR-0041) (#781)1Docsb607d7d2026-10-27 20:40
Change detail
1 change across 1 operation
- GET/v1/analytics
- info
endpoint added
- info
- feat(agents): deleting an agent archives it, so the runs it made still show it (#799)1Docs726fac52026-10-22 15:59
- fix(wallet): the wallet and payments routes take the agent's slug as well as its UUID (#798)7Docsece44da2026-10-22 15:53
- feat(authz): a run acts as its agent; saved tasks apart from triggers (ADR-0046 phase 1) (#777)728Docsf0235752026-10-20 15:41
- feat(skills): the repin moves installs holding a curated stub to the item's real folder (#775)1Docs7ac2c442026-10-20 11:12
- feat(skills): refresh-stub-installs re-reads a skill holding an old index's stub from GitHub (#771)1Docsef9c46d2026-10-20 09:42
- feat(skills): workspace skills as content-addressed packages with version history and rollback (#767)2Docs89986212026-10-20 09:25
- fix(registries): reconcile applies the whole manifest entry; platform credentials by reference; remove RegistrySync and sync_mode (#763)82Docs0fb1efa2026-10-20 08:36
- feat(skills): SKILL.md read at activation into the task log; runs wait out storage outages; install policy holds for installed skills at run time (#762)1Docs437a3552026-10-20 03:16
- fix(oauth): only the person who started a sign-in can finish it (#715) (#764)1Docs2d5a28d2026-10-20 01:06
- feat(events): event redesign — broker transport, task log, providers as catalog data, MCP Events in and out, platform stream, processors, trigger batches, workspace events feed (ADR-0040/42/43) (#748)4698155Docs8527c152026-10-19 21:57
- feat(skills): skill registries — sources, pinned installs, content by digest (ADR-0044) (#749)97Docsb8247902026-10-19 16:33
- feat(costs): tool-call cost metering (ADR-0038), task context and webhooks, stop/retry, scoped spend caps, catalog pricing (#739)995Docs221794d2026-10-13 19:04
- AgentArea organization: spend ledger + per-agent monthly cap, tool-call approvals, OpenFGA model in the image, reconcile cleanup (#737)9Docs04a1b802026-10-12 13:12
- feat(oauth): one connect flow and one callback for MCP servers and catalog APIs (#736)68Docs6ba34862026-10-12 10:03
- feat(access): a Telegram bot answers only linked people allowed to run its agent (#734)3Docs05995e52026-10-10 10:34
- fix(streams): a stream that triggers or forwards depend on cannot be deleted (#732)1Docsf1718bd2026-10-09 19:55
- fix(api): validate admin endpoint input, idempotent grants, body limit for chunked uploads (#730)33Docs477c5982026-10-09 18:12
- fix(files): restoring a file archives the one at its path instead of overwriting it (#729)1Docs2ff92f72026-10-09 17:52
- fix(skills,bundles): refuse bad input before anything is stored (#726)313Docs647237d2026-10-09 16:56
- fix(wallet): refuse wallet types, statuses and periods the wallet does not know (#701)59Docsb2370032026-10-09 14:06
- fix(governance): a policy priority outside the int4 range is a 422, not a 500 (#698)6Docs7fa81b02026-10-09 13:28
- fix: next 16.3.8, malformed-id 422, --repair-ownership wiring, OpenAPI last call + configured query params (#687)4Docs02e365e2026-10-09 10:51
- fix(connections): catalog API connections use the operator's OAuth apps; reconcile can retire a registry (#672)12Docsb4d18892026-10-09 10:32
- Next: stream connectors, graph grant ordering, and in-flight fixes (#667)11Docsd0ff7c02026-10-08 09:48
- feat(streams): event streams stage 1 — journal, fan-out subscriptions, Events page (ADR-0037) (#649)156Docs856ee402026-10-06 15:35
- feat(catalog): informative connection detail + 'where it runs' facet + full-width detail (#643)8Docs2649d562026-10-05 10:55
- feat(workspaces): per-workspace logo (REST, settings UI, MCP update) (#640)4Docsafe05bf2026-10-05 06:57
- Connect OAuth-only MCP servers through the operator's platform app; instance transport column; bundle as one plan (#639)5Docs34ef9b32026-10-05 00:47
- Release candidate: billing correctness, governance, approvals, signed webhooks, audit (#633)23225Docsa1fa0f02026-10-03 14:58
- feat: model kinds with media and decision tools, a wait tool, LLM trigger conditions and client toolsets (#618)1675Docse72e3aa2026-10-02 00:45
- feat(a2a): agent hosts by the spec, delegation in settings, cross-account agent keys (#620)21Docs881c7b52026-10-02 00:23
- test(e2e): make the real-stack release gate measure the product again (#602)1Docs604a3a72026-09-30 23:49
- fix(execution): make day-long agent runs survive their own length and infrastructure failures (#590)3Docs5a249682026-09-30 18:36
- feat(openapi): secret URL variables so an API key can live in the path (#577)6Docsf46ed472026-09-28 22:34
- feat(files): upload every file straight to storage with presigned URLs (#572)23Docsd8d73d82026-09-28 16:24
- feat(platform): record LLM cost in the billing currency via a customer_pricing extension (#536)2Docsc4f1b502026-09-27 21:12
- feat(files): sync a folder into workspace storage, uploading only what changed (#565)1Docs68456382026-09-27 20:43