v0.3.4

What's Changed

• feat(evals): add the sub-agent and modules explicit-request cases by @victorludvig in #349
• feat(evals): add the single-file explicit-request cases f02–f09 by @victorludvig in #354
• feat(evals): add input-level cases l3 and l4 by @victorludvig in #350
• test(evals): lint every case's own starting project by @victorludvig in #358
• fix(sdk): accept agent_usage as a constraint root in policy lint by @guillaume-hexamind in #363
• feat(evals): hold out four cases and add a model task parameter by @victorludvig in #364
• build(evals): cap parallel eval sandboxes at four by default by @victorludvig in #365
• feat(evals): route the Claude sandbox through an Anthropic-only egress proxy by @victorludvig in #355
• feat(evals): add must-refuse and robustness cases by @victorludvig in #366
• feat(evals): add security cases by @victorludvig in #367
• refactor(sdk): pass usage roots to policy engines as one decision input by @guillaume-hexamind in #369
• docs(sdk): fail on undocumented sdk surface and close the gaps by @quanghexa94 in #347
• fix(sdk): build the enforcer's usage source and audit sender from the platform client by @guillaume-hexamind in #371
• feat(platform-api): sign oauth access tokens with their own keypair by @victorludvig in #332
• feat(clickhouse,platform-api): roll usage up per end user and serve it by @guillaume-hexamind in #370
• feat(mcp): cache the api's jwks for the mcp server by @victorludvig in #333
• demo: 3-level DevOps deploy-chain showcase by @quanghexa94 in #360
• refactor(sdk): key usage ledgers and snapshots by usage scope by @guillaume-hexamind in #372
• feat(platform-api): fetch oauth client metadata behind an ssrf guard by @victorludvig in #334
• feat(platform-api): rank a project's end users by usage for the dashboard by @guillaume-hexamind in #373
• feat(dashboard): playground presentation mode by @quanghexa94 in #361
• feat(sdk): count usage per end user in bounded process-local ledgers by @guillaume-hexamind in #374
• feat(mcp): verify oauth tokens and serve the mcp app at /mcp by @victorludvig in #376
• feat(platform-api): cache oauth client metadata and match redirect uris by @victorludvig in #377
• feat(mcp): serve the policy reference and send server instructions by @victorludvig in #386
• feat(sdk): cap one end user's usage across agents with user_usage.* by @guillaume-hexamind in #378
• feat(sdk): read end-user usage from the platform, bounded per process by @guillaume-hexamind in #379
• feat(platform-api): serve oauth discovery metadata and jwks at the root by @victorludvig in #387
• fix(mcp): read the issuer from HEXGATE_PUBLIC_URL only and serve it unchanged by @victorludvig in #396
• build(mcp): build the mcp server image and check it in ci by @victorludvig in #397
• build: cap the project at python 3.13 (<3.14) by @quanghexa94 in #329
• feat(sdk): enforce user usage across processes at the run boundary by @guillaume-hexamind in #398
• feat(mcp): call the platform api as the caller and list their granted projects by @victorludvig in #390
• feat(platform-api): accept oauth access tokens in require_user on scoped routes only by @victorludvig in #403

Full Changelog: v0.3.3...v0.3.4

v0.3.3

What's Changed

• feat(sdk): surface agent guards on the manifest for platform governance by @quanghexa94 in #283
• feat(sdk): add the guards policy block to enable/disable declared guards by @quanghexa94 in #298
• feat(sdk): apply the guard policy stance in the guarded runner, closed-world by @quanghexa94 in #285
• feat(sdk): lint the guards policy block against the agent manifest by @quanghexa94 in #286
• docs(sdk): add a guarded-agent platform demo notebook by @quanghexa94 in #289
• docs: remove committed agent-policy-dimension design doc by @quanghexa94 in #277
• feat(sdk): add a skills block to the compose policy.yaml grammar by @guillaume-hexamind in #302
• feat(sdk): run every policy check through one entry point by @victorludvig in #306
• feat(clickhouse): record an agent_run row for every admitted run by @guillaume-hexamind in #310
• feat(sdk): add a process-local per-agent usage ledger by @guillaume-hexamind in #308
• feat(platform-api): route policy preview, check and validate through analyze_policy by @victorludvig in #304
• refactor(cli): route policy validate through analyze_policy by @victorludvig in #318
• feat(sdk): check gate args, skill names and constraint roots in analyze_policy by @victorludvig in #321
• refactor(sdk): fetch policy and bans concurrently at the run boundary by @guillaume-hexamind in #309
• feat(sdk): lint each agent's roles column against its own manifest by @victorludvig in #322
• feat(sdk): flag reach targets and roles columns that name no agent by @victorludvig in #324
• feat(clickhouse): roll usage up per agent per minute, deduped across polls by @guillaume-hexamind in #317
• feat(evals): score a policy-writing workspace with the hexgate sdk by @victorludvig in #287
• feat(platform-api): add the oauth scope registry and token tables by @victorludvig in #327
• fix(sdk): canonicalize the native agent's name at every identity site by @guillaume-hexamind in #325
• feat(evals): score role-wide cases on every roles.yaml column (land #299) by @victorludvig in #337
• feat(sdk): make agent_usage.* paths loadable, read from the local ledger by @guillaume-hexamind in #314
• feat(platform-api): serve agent usage windows from the minute rollup by @guillaume-hexamind in #326
• feat(evals): load the policy-writing cases, one directory per case by @victorludvig in #288
• feat(evals): check a case's dry-run calls against its agent by @victorludvig in #330
• feat(evals): add the single-file starting project and the first case by @victorludvig in #339
• feat(evals): run the policy-writing cases in Inspect on our own answers by @victorludvig in #342
• feat(sdk): add the platform usage source behind agent_usage.* by @guillaume-hexamind in #331
• feat(sdk): add the usage.on_unavailable policy key, signed into the bundle by @guillaume-hexamind in #335
• feat(evals): run Claude Code in a Docker sandbox for agent=claude by @victorludvig in #343
• build(evals): install OPA in the Claude sandbox image per architecture by @victorludvig in #345
• feat(sdk): enforce agent_usage.* across processes at the run boundary by @guillaume-hexamind in #338
• feat(evals): add input-level cases l0, l1 and l5 by @victorludvig in #348
• feat(evals): add the sub-agent and modules starting projects by @victorludvig in #341
• feat(evals): add input-level cases l2 and l2b by @victorludvig in #352
• docs(sdk): document agent_usage.* as a soft cross-process quota by @guillaume-hexamind in #340
• fix(sdk): percent-encode the agent name in get_agent's url by @Dev-next-gen in #359

New Contributors

• @Dev-next-gen made their first contribution in #359

Full Changelog: v0.3.2...v0.3.3

v0.3.2

What's Changed

• fix(sdk, collector): apply role-keyed top-level constraints, widen revocation staleness by @guillaume-hexamind in #199
• feat(platform-api): compose policy-file store + resolve/check/preview API by @quanghexa94 in #179
• style(platform-api): ruff-format policy_modules service/router/tests by @quanghexa94 in #207
• feat(sdk): compose admission block (role→agent ingress access) by @quanghexa94 in #200
• feat(clickhouse): add llm_message table by @victorludvig in #183
• fix(redpanda): raise the record-size budget for message spans by @victorludvig in #185
• feat(platform-api, collector): soft-delete devtoken rows and filter revoked keys on every read path by @guillaume-hexamind in #206
• feat(platform-api): map hexgate.messages spans into LlmMessageEvent by @victorludvig in #186
• feat(sdk): add LlmMessageEvent and emit_llm_messages by @victorludvig in #187
• feat(platform-api): store llm_message rows from the enricher by @victorludvig in #188
• feat(sdk): add the MessageCursor delta helper by @victorludvig in #189
• feat(platform-api, dashboard): record who created and last updated every control-plane row by @guillaume-hexamind in #209
• feat(dashboard): compose policy editor (file store, live preview, graph, tester) by @quanghexa94 in #180
• feat(platform-api): add the session-scoped llm-messages read endpoint by @victorludvig in #190
• feat(dashboard): move org + account admin into a settings space by @quanghexa94 in #219
• feat(dashboard): theme system — light/dark modes and blue/plum accent by @quanghexa94 in #218
• feat(sdk): capture OpenAI Agents messages in HexgateUsageHooks by @victorludvig in #191
• fix(platform-api): union policy graph to the most-permissive verdict by @quanghexa94 in #222
• refactor(sdk): enable ruff isort and sort imports tree-wide by @quanghexa94 in #145
• test(platform-scripts): cover hexgate.messages in the OTLP smoke check by @victorludvig in #192
• fix(dashboard): resolved-tab agent dropdown, inline folder create, legible graph edges by @quanghexa94 in #223
• fix(platform-scripts): apply each store's migrations independently by @guillaume-hexamind in #220
• feat(sdk): capture LangChain messages in HexgateUsageCallbackHandler by @victorludvig in #193
• fix(platform-scripts): guard each migration against a table create_all has not built by @guillaume-hexamind in #224
• build(platform-api): rebuild the SDK wheel on every image build by @guillaume-hexamind in #227
• feat(dashboard): show LLM messages in the Audit decision drawer by @victorludvig in #195
• feat(sdk): capture Google ADK messages in HexgateUsagePlugin by @victorludvig in #194
• feat(sdk): emit a per-run message event for Pydantic AI by @victorludvig in #196
• feat(dashboard): modernize the playground into an agentic chat by @quanghexa94 in #226
• test(platform-api): execute the migrations in CI against real Postgres and ClickHouse by @guillaume-hexamind in #228
• feat(sdk): tabulate a resolved policy set as an authorisation matrix by @victorludvig in #238
• feat(platform-api): record the operator's AI Act classification per agent by @victorludvig in #235
• feat(platform-api): assemble and sign the AI Act evidence report by @victorludvig in #239
• feat(sdk): add skill definitions to the agent manifest by @guillaume-hexamind in #253
• fix(sdk): expand Google ADK toolsets in the manifest and policy gate by @guillaume-hexamind in #254
• feat(sdk): child.as_tool() for native agent-as-tool sub-agent reach by @quanghexa94 in #233
• feat(sdk): enumerate_subagents — per-framework sub-agent introspection by @quanghexa94 in #243
• feat(sdk): record sub-agent reach edges on the manifest + via-specific warn by @quanghexa94 in #244
• feat(cli): register the whole sub-agent tree (register_tree) + serve/register flags by @quanghexa94 in #245
• docs: add pr-description and write-adr agent skills by @quanghexa94 in #217
• feat(platform-api): persist agent skills from the manifest by @guillaume-hexamind in #256
• feat(dashboard): show registered skills on the agents page by @guillaume-hexamind in #258
• feat(sdk): discover Google ADK skills at registration by @guillaume-hexamind in #259
• feat(platform-api): render the AI Act evidence annex to PDF by @victorludvig in #237
• chore(skills): add write-policy Claude skill by @victorludvig in #263
• test(sdk): probe the ADK skills contract in the version matrix by @guillaume-hexamind in #261
• feat(sdk): add the skill policy key namespace by @guillaume-hexamind in #262
• chore(skills): add run-platform skill by @victorludvig in #265
• feat(cli): bind sub-agents to platform policy at serve by @quanghexa94 in #271
• docs(repo): rename CLAUDE.md to AGENTS.md by @victorludvig in #274
• feat(dashboard): ai act evidence report tab by @victorludvig in #236
• feat(sdk): evaluate skill keys closed-world on both engines by @guillaume-hexamind in #266
• feat(sdk): gate Google ADK skill activation under skill policy keys by @guillaume-hexamind in #267
• feat(platform-api): generate a skills block in the starter policy by @guillaume-hexamind in #270
• fix(sdk): discover the real tool surface of a compiled LangGraph agent by @guillaume-hexamind in #272
• test(sdk): assert the deepagents probe gates the top-level bound tool surface by @guillaume-hexamind in #275
• feat(sdk): discover deepagents skills at registration by @guillaume-hexamind in #278
• feat(sdk): gate deepagents skill activation under skill policy keys by @guillaume-hexamind in #282
• demo: support_bot + billing_bot policy showcase by @quanghexa94 in #264
• test(sdk): probe the deepagents skills contract in the version matrix by @guillaume-hexamind in #291
• feat(platform-scripts): seed llm_message transcripts alongside decisions by @victorludvig in #230
• fix(sdk): keep the cursor mark on an empty list and skip only a missing prompt by @victorludvig in #229
• docs(sdk): document agent skills, their policy block and both adapters by @guillaume-hexamind in #292
• fix(platform-api): install weasyprint system deps in the release job by @guillaume-hexamind in #296

Full Changelog: v0.3.1...v0.3.2

v0.3.1

What's Changed

• refactor(dashboard): rename tokens page to API keys, tidy nav and copy by @guillaume-hexamind in #102
• refactor(platform-api): rename DevToken to ApiKey by @victorludvig in #108
• build(collector): scaffold the OTLP ingestion Collector (Go) by @victorludvig in #111
• docs(sdk): guards + official-plugins concept page by @quanghexa94 in #125
• fix(platform-api): make the signed token_id fact the api-key row id by @victorludvig in #126
• feat(collector): verify Hexgate API keys in the OTLP receiver by @victorludvig in #127
• feat(sdk): agent-level policy model with admission/agents blocks by @quanghexa94 in #122
• feat(sdk): official secret-detection guards (hexgate.plugins) by @quanghexa94 in #123
• feat(platform-api): add ClickHouse batch-insert functions for the enricher job by @victorludvig in #128
• feat(sdk): define otel span semconv, expose redaction pipeline by @victorludvig in #130
• fix(platform-api): cover llm_invocation in the startup schema check by @victorludvig in #131
• refactor(platform-api): aggregate schema gaps across features at startup by @victorludvig in #136
• feat(sdk): serve OpenAI, Google ADK, and Pydantic AI agents (framework-agnostic streaming) by @quanghexa94 in #147
• Revert "feat(sdk): serve OpenAI, Google ADK, and Pydantic AI agents (framework-agnostic streaming)" by @quanghexa94 in #148
• feat(platform-api): add the span-enricher job (redpanda to clickhouse) by @victorludvig in #133
• feat(sdk): RunFacts accumulator and run scope by @guillaume-hexamind in #129
• feat(platform-api): compile modular agents' bundles from resolved modules by @quanghexa94 in #121
• feat(sdk): serve OpenAI, Google ADK, and Pydantic AI agents (framework-agnostic streaming) by @quanghexa94 in #149
• test(sdk): pin run scope coverage for the streamed launch helper by @victorludvig in #163
• fix(platform-scripts): realign seed_audit with the audit column contract by @guillaume-hexamind in #159
• feat(sdk): emit audit events as otlp spans by @victorludvig in #146
• build(collector): ship the full otlp pipeline in the deploy stack by @victorludvig in #157
• build(platform-scripts): add the otlp pipeline smoke test as a post-deploy check by @victorludvig in #168
• feat(sdk): agent-level admission enforcement (AgentGate, closed-world agent keys) by @quanghexa94 in #124
• docs(platform-api): describe the otlp pipeline as the audit transport by @victorludvig in #170
• fix(platform-api): raise audit log retention to six months by @victorludvig in #174
• feat(sdk): run namespace and writer by @guillaume-hexamind in #150
• feat(sdk): compose policy.yaml grammar front-end by @quanghexa94 in #175
• feat(clickhouse, platform-api): run_id and run counters on policy_decision by @guillaume-hexamind in #153
• docs(sdk): marimo notebook demoing guards + plugins across frameworks by @quanghexa94 in #156
• feat(sdk): enforce agent reach + admission at runtime, cap handoff depth (A3+A4) by @quanghexa94 in #142
• feat(sdk): add hexgate.messages wire contract and content caps by @victorludvig in #182
• feat(sdk): compose import graph (position=scope imports) by @quanghexa94 in #176
• feat(sdk): run facts on the audit payload, and policy-level constraints by @guillaume-hexamind in #155
• refactor(dashboard): move workspace switcher into a collapsible sidebar by @quanghexa94 in #172

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

• tests(sdk): find the framework versions compatible with hexgate adapters by @guillaume-hexamind in #91
• fix(sdk): correct framework version support and publish compatible ranges by @guillaume-hexamind in #92
• docs(sdk): unify adapter examples around one devops agent by @guillaume-hexamind in #94
• fix(hexgate): pydantic ai run sync no loop drop by @victorludvig in #93
• feat(sdk): optional egress proxy — gate outbound network by host by @quanghexa94 in #95
• feat(sdk): raw-TCP egress reachability gate (gate DB connections by host:port) by @quanghexa94 in #96
• docs(sdk): replace readme decision-flow ascii diagram with svg by @victorludvig in #98
• feat(sdk): multi-module policy — modules + linker (1/3) by @quanghexa94 in #97
• test(sdk): add enforcement + policy-pull latency benchmarks by @guillaume-hexamind in #100
• test(sdk): add live integration tests for all four framework adapters by @victorludvig in #99
• build(claude): add integration test skill by @victorludvig in #104
• fix(sdk): fail open when emit_llm_usage raises in adapter usage hooks by @victorludvig in #105
• feat(sdk): replace User scope with HexgateContext (user_roles + attributes) by @guillaume-hexamind in #106
• feat(sdk): thread ctx.* attributes through both policy engines by @guillaume-hexamind in #107
• feat(sdk): policy analyzer + hexgate policy check (2/3) by @quanghexa94 in #101
• refactor(sdk): user -> context/hexgate_context rename + ctx.* docs by @guillaume-hexamind in #112
• build(redpanda): stand up local Redpanda infra for OTLP ingestion by @victorludvig in #110
• refactor(sdk): user -> context rename minor fixes by @guillaume-hexamind in #114
• feat(sdk, platform): send attributes in audit events, update clickhouse schema by @guillaume-hexamind in #113
• feat(sdk): role-aware policy assembly (roles as capability bindings) by @quanghexa94 in #115
• feat(sdk): multi-role policy enforcement by @guillaume-hexamind in #117
• feat(sdk): pre/post tool-hook pipeline with observe, halt, and arg rewrite by @quanghexa94 in #118
• feat(sdk): route OpenAI/Google/Pydantic AI adapters through run_guarded (guards on all frameworks) by @quanghexa94 in #119
• feat(platform-api): project policy-module store + resolve/check API by @quanghexa94 in #116
• feat(sdk, platform): persist multi-role decisions in the audit pipeline by @guillaume-hexamind in #120

Full Changelog: v0.2.11...v0.3.0

v0.2.11

What's Changed

• fix(sdk): resolve .env from consumer cwd, not installed package path by @guillaume-hexamind in #85
• docs(sdk): add public ban concepts page by @guillaume-hexamind in #86
• docs(sdk): fix readme badge row spacing by @victorludvig in #89
• Add the gates demo: marimo narrative, policy, and one-box launch by @quanghexa94 in #81
• build(claude): add per-file format-on-edit hook by @victorludvig in #88
• feat(platform-scripts): enhance seed_audit.py script by @victorludvig in #84
• docs: remove legacy documentation, trim and optimize documentation not sdk and platform consumer - readable by @guillaume-hexamind in #87
• feat(dashboard): link key pages to hexgate docs by @guillaume-hexamind in #90

Full Changelog: v0.2.10...v0.2.11

v0.2.10

What's Changed

• feat(dashboard): usage dashboard page by @victorludvig in #83

Full Changelog: v0.2.9...v0.2.10

v0.2.9

What's Changed

• feat(hexgate): add llm usage hook by @victorludvig in #76
• refactor(sdk): split approval + tracing types so adapters skip LangChain by @quanghexa94 in #80
• feat(platform-api): add llm invocation summary API endpoint by @victorludvig in #82
• feat(demo): live demo on Daytona + Modal spawner (retire Codespaces) by @quanghexa94 in #78
• feat(dashboard): agent & user bans dashboard by @guillaume-hexamind in #79

Full Changelog: v0.2.8...v0.2.9

v0.2.8

What's Changed

• refactor(sdk): slim core - remove demo content, fix layering & naming by @guillaume-hexamind in #66
• fix(dashboard): Graph reads registered manifest as source of tools by @quanghexa94 in #68
• fix(platform-api): patch clickhouse_ping at its actual import site in readiness test by @victorludvig in #71
• feat(platform-api): Create llm invocation audit table and ingest POST endpoint by @victorludvig in #72
• feat(sdk): expand constraint DSL + policy-in-code, with pydantic↔WASM parity by @quanghexa94 in #69
• docs: README and documentation restructuration by @guillaume-hexamind in #70
• refactor(sdk): Generalize audit api keys registry by @victorludvig in #73
• feat(platform/api): ban invokation api backend logic by @guillaume-hexamind in #74
• docs: marimo demo for gating an MCP server with policy by @quanghexa94 in #75
• feat(hexgate): Agent or User Ban Gate by @guillaume-hexamind in #77

Full Changelog: v0.2.7...v0.2.8

v0.2.7

What's Changed

• refactor(platform/api): carve flat modules into core/deps/domains packages by @guillaume-hexamind in #64
• feat(playground): route approval_required tool calls to the connected tab by @quanghexa94 in #63
• feat(mcp): fan out MCP tool wrapping to every adapter by @quanghexa94 in #65

Full Changelog: v0.2.6...v0.2.7