• Give a judge case the iteration it graded (#4477)latestOpenAPI 3.1.0○6bb1e0972026-08-29 01:56
• Evals D5b + D5c: serve per-run stage analytics, and show where the chain stopped (#4455)OpenAPI 3.1.0○151294e72026-08-28 20:36
• feat(evals): wire B5c intent and stage measurements (#4441)OpenAPI 3.1.0○840256452026-08-27 23:58
• Versioned Skills: surface skill revisions in run comparisons and the environment picker (#4431)OpenAPI 3.1.0○1218e3f0f2026-08-27 22:10
• Skills: review follow-ups from #4433, and two docs claims that stopped being true (#4437)OpenAPI 3.1.0▲2○3160164b52026-08-27 21:50
• Skills over MCP: inspection on every agent surface (#4433)OpenAPI 3.1.0○3acb49ed2026-08-27 21:36
• Skills: unbreak harness eval runs, and make skill IDs obtainable (#4409)OpenAPI 3.1.0○2fbe5a242026-08-27 04:40
• evals(Lane F): an imported case carries its claim, and an approximation is approved per run (#4391)OpenAPI 3.1.0●1○1358f90a32026-08-26 20:53
• Agent-emitted permalinks: every durable resource hands back the app URL that owns it (#4375)OpenAPI 3.1.0○117b7afe2026-08-26 05:44
• clients: agent surfaces can edit clients, in the product's own vocabulary (#4378)OpenAPI 3.1.0▲7○15a334fc52026-08-26 05:19
• Evals D9: one canonical run decision summary across API, MCP and CLI (#4374)OpenAPI 3.1.0○2b26dfa72026-08-26 00:42
• evals(E4b): pin an eval baseline by source commit SHA (#4370)OpenAPI 3.1.0●1○4560d86f2026-08-25 23:21
• evals(E5b): gate waivers end to end — a waived gate is never a silent pass (#4371)OpenAPI 3.1.0○685f8e112026-08-25 23:09
• Evals v2 G4c: disclose the host axis before a run launches (#4350)OpenAPI 3.1.0○18c74ba22026-08-25 03:35
• Evals v2 — G4b: the inspector half of pre-run disclosure (#4334)OpenAPI 3.1.0○189d43272026-08-25 00:19
• evals: run under the v2 verdict policy end to end (B4 Phase 2) (#4332)OpenAPI 3.1.0●6○37ed76d0b2026-08-24 21:12
• feat(widgets): let agents see and drive MCP App widgets (#4312)OpenAPI 3.1.0○10a416fc2026-08-24 04:12
• feat(api): agent Playground — drive a session from API/CLI/MCP and read its telemetry (#4309)OpenAPI 3.1.0○37f45c122026-08-24 01:03
• eval: evidence refs and decision summary for failed cases (#4307)OpenAPI 3.1.0○55ffc2cf2026-08-23 22:44
• v1: add durationMs to tools/call (#4292)OpenAPI 3.1.0○1b20453c2026-08-23 21:38
• eval: add `eval run --file` for versioned suite files (#4294)OpenAPI 3.1.0○53b0cc1a2026-08-23 04:08
• eval: variadic compose-model, ephemeral launch, ad-hoc id passthrough (#4261)OpenAPI 3.1.0○342e71852026-08-22 23:32
• eval: client × model composer fan-out (#4256)OpenAPI 3.1.0●1○153d0382e2026-08-22 23:31
• feat: add hosted conformance start and poll agent ops (#4272)OpenAPI 3.1.0○4fb61fa22026-08-22 23:27
• I1: Registry v1 proxies, install adapters, and SDK client (#4251)OpenAPI 3.1.0○89abc30e2026-08-22 22:44
• cli: identify local MCP as MCPJam CLI (#4203)OpenAPI 3.1.0no API changes4e6208f2026-08-22 04:10
• Reject unknown v1 eval/host write-body keys with a named 400 (#4226)OpenAPI 3.1.0▲1○6ae21e092026-08-22 02:20
• Durable conformance run history, CLI upload, and GitHub dual-check worker (#4197)OpenAPI 3.1.0○59bd21c02026-08-21 18:32
• Directory readiness: bound the archive read, honest replay receipts, deadlined report reads (#4158)OpenAPI 3.1.0●24○35a31cb4c2026-08-20 21:22
• Directory readiness on every product surface: the SDK algebra, the worker, and the surfaces (#4153)OpenAPI 3.1.0○673f1c2d2026-08-20 20:58
• fix(docs): dedupe the OpenAPI keys that have failed every Mintlify build since Aug 16 (#4136)OpenAPI 3.1.0▲7●3○945aedab2026-08-19 23:48
• evals: an idempotent retry replays its run instead of running it again (#4134)OpenAPI 3.1.0▲1○25f88e912026-08-19 23:33
• CLI/MCP parity for eval-run combinations: explicit fan-out, atomic batch launch, composed targets, harness honesty (#4124)OpenAPI 3.1.0○155b5d7842026-08-19 22:45
• Suite settings on the agent surfaces (MCP / CLI / SDK) (#4119)OpenAPI 3.1.0○1815c3bf32026-08-19 21:20
• Evals W0.3b: batch case authoring, declared ids, and a dual-shape generation reader (#4112)OpenAPI 3.1.0○7e6f57142026-08-19 01:12
• Evals v2 G1e: publish the eval role matrix as capabilities, and gate deletes on it (#4109)OpenAPI 3.1.0○68d8903b2026-08-18 23:37
• Rename chatbox to scenario (#4050)OpenAPI 3.1.0▲2●3○1043bd0882026-08-16 20:53
• Evals Tranches 2–4: run comparison + comparative gates, hosted corpus lock, @mcpjam/vitest (#4048)OpenAPI 3.1.0○1dd212de2026-08-15 19:15
• search_sessions: one search operation, adapted per surface (#4049)OpenAPI 3.1.0○1db2b0662026-08-15 19:13
• Launch guardrails: 429 transport + gateway limiter, and the OpenAPI A-chain (#4045)OpenAPI 3.1.0▲15●8○5855317612026-08-15 09:36
• Evals: one versioned evaluation contract (Tranche 1) (#4044)OpenAPI 3.1.0○6f89e97e2026-08-15 08:55
• feat(platform): organizations read end-to-end, project create/update on the MCP catalog (#3970)OpenAPI 3.1.0○1c7c10dc2026-08-14 01:54
• feat(server-connections): worker runtime, REST + handoff routes, SDK operations, CLI (#3958)OpenAPI 3.1.0○4bb0346d2026-08-13 23:33
• Canonical provider classification, forward-client model invariant, and the environment model on the public surface (#3936)OpenAPI 3.1.0▲1○12e7c8d622026-08-13 21:17
• feat(plugins): chatbox plugin provenance, worker plugin read tools, httpVariant transport mapping (Track A) (#3841)OpenAPI 3.1.0○295e303f2026-08-10 16:08
• Surface-map program: auth fixes, public journey/scenario API, Discord correctness, drift ratchets, docs (#3791)OpenAPI 3.1.0○15f31eeb02026-08-09 22:35
• Cut over Slack integration settings to the org-level Slack bot (PR-I1) (#3781)OpenAPI 3.1.0no API changesba793992026-08-08 00:12
• [codex] complete platform bootstrap and eval parity (#3735)OpenAPI 3.1.0○11c47526c2026-08-06 05:51
• Agent approval hardening: unforgeable preview, one oracle-free answer, targeted dedup, honest evidence (#3707)OpenAPI 3.1.0○101515862026-08-04 07:10
• feat(evals): make the SDK/CLI/MCP speak environments (#3693)OpenAPI 3.1.0▲1○4b84413a2026-08-04 07:05

Show older