PostHog

Account Notes List

Account Relationship Definitions Create

Account Relationship Definitions Destroy

Account Relationship Definitions List

Account Relationship Definitions Partial Update

Account Relationship Definitions Retrieve

Account Relationship Definitions Update

Accounts Create

Accounts Custom Property Values Create

Accounts Custom Property Values List

Accounts Destroy

Accounts List

Accounts Notebooks Create

Accounts Notebooks Destroy

Accounts Notebooks List

Accounts Notebooks Retrieve

Accounts Partial Update

Accounts Relationships Create

Accounts Relationships End Create

Accounts Relationships List

Accounts Retrieve

Accounts Update

Bulk update tags on multiple objects. PAT access: this action has no ``required_scopes=`` on the decorator — inheriting viewsets must add ``"bulk_update_tags"`` to their ``scope_object_write_actions`` list to accept personal API keys. Without that opt-in, ``APIScopePermission`` rejects PAT requests with "This action does not support personal API key access". Done per-viewset so granting ``<scope>:write`` for one resource doesn't leak access to sibling resources that share this mixin. Accepts: - {"ids": [...], "action": "add"|"remove"|"set", "tags": ["tag1", "tag2"]} Actions: - "add": Add tags to existing tags on each object - "remove": Remove specific tags from each object - "set": Replace all tags on each object with the provided list

Actions Create

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Actions List

Actions Partial Update

Actions References List

Actions Retrieve

Actions Update

Activity Log List

Advanced Activity Logs Available Filters Retrieve

Advanced Activity Logs Export Create

Advanced Activity Logs List

Approve or reject a queued `agent`-type tool-approval request. This is the OWNER decision surface — the only PostHog-authoritative one: team admins decide here, in the console. `principal`-type approvals are decided by the session principal at the ingress decision API, not here. The runtime side runs the tool platform-side on approve and wakes the session with a synthetic tool_result either way.

List approval-gated tool requests for this application. Team-admin only (per plan §6.1). Default returns all states — pass `?state=queued` for the inbox view.

Single approval request — full proposed args, assistant snapshot, decision metadata, dispatch outcome. Team-admin only (plan §6.1).

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Served-model catalog — each model's id, provider, context window, and USD-per-million-token pricing — plus the curated auto-level → model map. Project-agnostic; sourced from the AI gateway catalog. Powers the config UI model browser and the agent builder's model-choosing skill.

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Authoring-side proxy for invoking a *draft* (or any non-live) revision. Closes the anonymous-draft-invoke gap: the public ingress URL refuses non-live invokes that don't carry the `x-agent-preview-secret` header; this proxy attaches it after authenticating the Django caller. URL: `/api/projects/<team>/agent_applications/<app>/preview-proxy/<rest>` Auth: standard PAT / session — `agents:write` scope (POST run/send/cancel is a mutating invoke; the read-only `listen` GET is `agents:read`).

GET passthrough for the preview-proxy — used for `/listen` SSE.

GET sibling of `preview_token_mint`. Same body and response shape — exists because `EventSource` can't set headers, so SSE callers fetch the token via GET and then attach `?preview_token=` to the ingress URL. Behind the same URL (`url_path="preview-token"`) thanks to DRF's `@<action>.mapping.get`; DRF resolves it to a distinct `view.action`, but it is in `scope_object_write_actions` alongside the POST sibling — both return a usable credential, so both require `agents:write`.

Mint a short-lived JWT for talking to a non-live revision directly via the public ingress URL. The caller attaches it as the `x-agent-preview-token` header (or `?preview_token=` query param for `EventSource`). See `_mint_preview_jwt` for the payload + claim binding. The response also includes `endpoints`, `auth`, and `preview_proxy` blocks so the caller can wire a preview invocation without grepping the agent-ingress source for which path each trigger exposes or which header name carries the token. This is the "self-describing" half of preview-mode — every piece of info you need to hit ingress is in one response. POST is the canonical verb — minting credentials for downstream `run`/`send`/`cancel` is a write-class capability. A GET sibling exists at the same URL for `EventSource` callers (which can't set headers); it is also write-scoped, since it returns the same token.

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Mark a revision archived. If it was the live one, clear the application's live_revision pointer (the app effectively has no deployable version until another revision is promoted).

Update one `.md` file on a draft revision. `agent.md` writes go to the draft bundle. `skills/<id>/SKILL.md` writes are store-backed — skills are materialized from the skill store at freeze, so the edit publishes a new version of the referenced store skill and re-pins the draft's `skill_refs` entry to it. `<id>` must be a ref alias on this revision; add new skills via `bundle/import/` or `skill_refs`. Tool source / schema editing is out of scope here — use the per-tool endpoints. Returns the updated revision so the caller can refresh in one round-trip.

Bulk-merge a set of `.md` files into a draft revision. Sets `agent_md` on the draft bundle if present. `skills[]` are store-backed and merge by `id`: an id already referenced by the draft publishes a new version of its store skill; an unreferenced id attaches the store skill of that name (publishing the payload's body to it), or creates it when no such skill exists — and each ref is (re-)pinned to the published version. Skills not mentioned are left alone, so the import is safe to retry. Draft-only; non-draft revisions return 409 untouched.

Read the full typed bundle: `{ agent_md, skills, tools, spec }`.

Full-replace the typed bundle. Anything not in the payload is deleted. Tool sources are AST-checked + esbuild-compiled by the janitor before any S3 writes.

Copy every file from `source_revision_id` into this revision.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Fire one cron job out-of-band — the same execution path the scheduler walks, but on demand. Authoring UX: the user iterates on a cron prompt by clicking 'Fire now' rather than waiting for the next scheduled firing. Without this, 'did my prompt do the right thing?' is unanswerable until the cron actually fires. Idempotent via `request_id`: repeat clicks with the same id resolve to the same session id rather than firing N times.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Freeze the bundle: draft → ready, stamps sha256 on the row. Django is a thin proxy here: resolve template refs into the bundle, ask the janitor to seal it (the janitor returns the sha + the spec it derived from the typed resources), then stamp the row. No `transaction.atomic()` — the janitor's freeze is idempotent (on retry it re-reads the existing `.frozen` marker + re-derives spec), so a partial failure here is recoverable by re-calling freeze, not by transactional rollback. Holding an atomic block across the janitor HTTP call previously deadlocked the agent_revision row against the janitor's spec write — that's moved off the janitor side as part of the same fix.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

List every file in this revision's bundle (path, size, sha256).

Create a fresh draft revision under `application_id` and seed it from `source_revision_id`. Saves the MCP one round-trip vs the explicit create + clone_from sequence.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

ready → live. Sets the parent application's live_revision.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Replace this revision's encrypted env block. The body is `{ "env": { "<KEY>": "<value>", ... } }`. The encrypted text is stored on `AgentRevision.encrypted_env`; the worker decrypts it at session start via the same Fernet schedule (see agent-shared/src/runtime/encryption.ts).

Full-replace the draft's store-skill references. They are resolved and materialized into the bundle at freeze, not here — this only records which skills (and pinned versions) the freeze should pull in.

Build a Slack app manifest for this revision's slack trigger. Deterministic: the OAuth scopes and bot event subscriptions are derived from the slack trigger config (`mention_only` / `auto_resume_threads` / `ack_reaction`) and the agent's Slack tools, so the manifest already subscribes to exactly the events the config needs. 400 if the revision has no slack trigger.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Return the fully-assembled system prompt for this revision. Authoring tools call this to preview what the model will actually see at session start — the platform framework preamble plus the bundle's `agent.md` plus the skills index. Useful for debugging author-vs-framework precedence conflicts and verifying `spec.framework_prompt.omit` overrides took effect.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Execute one persisted custom tool in a single-shot sandbox. Authoring loop's "test this tool" button. The tool's source must already be PUT (compiled.js is what runs); this just invokes it with the caller-supplied args and a stubbed ctx. No real secrets leave Django — `mock_secrets` is a `{name → placeholder}` map.

Revisions of an agent. Created in `draft`, promoted through `ready → live` once the bundle has been uploaded + frozen. URLs (nested under an application): Model CRUD: GET .../revisions/ list POST .../revisions/ create draft GET .../revisions/<id>/ retrieve PATCH .../revisions/<id>/ update spec (draft only) Lifecycle: POST .../revisions/<id>/promote/ ready → live POST .../revisions/<id>/archive/ → archived POST .../revisions/<id>/freeze/ draft → ready (stamps sha256) POST .../revisions/<id>/clone_from/ copy bundle from another rev POST .../revisions/new_draft/ create draft + clone_from atomically Bundle authoring (proxied to the janitor): GET .../revisions/<id>/manifest/ list paths + sha256 GET .../revisions/<id>/file/?path=… read one file PUT .../revisions/<id>/file/?path=… write one file (draft) DELETE .../revisions/<id>/file/?path=… delete one file (draft) GET .../revisions/<id>/bundle/ bulk pull all files PUT .../revisions/<id>/bundle/ bulk push (replace|merge)

Spec edits are only allowed while state='draft'. Once promoted to ready/live the spec is frozen — change requires a new revision.

Pre-flight checks before freeze + promote: agent.md exists, every native tool id is registered, every custom tool has its compiled.js + schema.json, every skill path exists, every declared secret has a value set in this revision's env block. Returns `{ ok, errors: [...] }`. Works on any revision state.

Read the runner's structured event log for one session from ClickHouse. Filters (limit / after / before / level / search) match the shared `LogEntryMixin` helper used by hog_function + hog_flow.

List sessions for this application, newest first. Strips the conversation transcript from each summary, but includes a `preview` (last assistant text, ~120 chars) and `usage_total` (token + cost aggregate). Use `agent-applications-sessions-retrieve` for the full transcript of a single session.

Fetch one session's state — full conversation by default, or just the trailing N messages with `?last_n=`. Always returns a `usage_total` block aggregated over the entire session, regardless of trim. The runner-side queue DB is the source of truth.

The canonical JSON Schema for an agent `spec` — every field, type, enum, default, and the discriminated unions for `models` / `triggers[]` / `tools[]`, each with an inline description. Emitted from the same source the runner validates against (fields with a default are optional on write), so read it BEFORE composing a spec for create / revisions-spec-update instead of guessing the shape. Pass `section` to fetch just one part.

Roll-up stats for the agent — drives the agent-detail overview tiles.

Agent applications — the deployable unit of the platform. URLs: GET /api/projects/<team>/agent_applications/ list POST /api/projects/<team>/agent_applications/ create GET /api/projects/<team>/agent_applications/<id|slug>/ retrieve PATCH /api/projects/<team>/agent_applications/<id|slug>/ update DELETE /api/projects/<team>/agent_applications/<id|slug>/ archive POST /api/projects/<team>/agent_applications/<id|slug>/set_env/ bulk replace env GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/ list set keys GET /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ is one key set? PUT /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ set one key DELETE /api/projects/<team>/agent_applications/<id|slug>/env_keys/<KEY>/ clear one key

Revoke one of an end-user's linked connections. The credential is marked revoked (kept for audit), so the agent can no longer act as that user on the provider.

List this agent's end-users (the stable identities behind inbound principals) and each user's linked external connections. Connection metadata only — credential material is never returned.

Approval-gated tool requests across every agent in this team. Team-admin only.

Live (non-terminal) sessions across every agent owned by this team, newest activity first.

Roll-up stats across every agent owned by this team.

Create a memory file. Fails if the path already exists — use the update endpoint to overwrite.

Hard-delete a memory file. Activity log captures the action against the agent.

Read one memory file in full (frontmatter + markdown body).

List memory file headers under the agent's prefix. Headers only — no bodies.

List the agent's tabular-reference tables (the @posthog/table-* JSONL tables): name + byte size.

Read rows from one tabular-reference table (capped via ?limit).

BM25 search across the agent's memory files. Ranks by description+tags+path+body with field weighting.

Pre-aggregated folder tree of memory files. Saves the frontend re-derivation work.

Update a memory file. Any field omitted is preserved from the existing file.

Read-only catalog of every @posthog/* native tool the runner knows.

GET / PUT / DELETE one secret by name on this revision. - `GET` → `{ key, is_set }` (never returns the value). - `PUT` → upserts `{ value }` into the env block. - `DELETE` → removes the key. No-op when it wasn't set. Per-method scope: GET is treated as a write action so the single action name maps to one consistent scope; reading whether a secret is set is restricted to writers in any case.

GET / PUT / DELETE one secret by name on this revision. - `GET` → `{ key, is_set }` (never returns the value). - `PUT` → upserts `{ value }` into the env block. - `DELETE` → removes the key. No-op when it wasn't set. Per-method scope: GET is treated as a write action so the single action name maps to one consistent scope; reading whether a secret is set is restricted to writers in any case.

List the names of secrets currently set on this revision. Returns names only — values stay server-side under `EncryptedTextField`. Use this to drive the "set / unset" badge next to a declared secret in the editor UI.

GET / PUT / DELETE one secret by name on this revision. - `GET` → `{ key, is_set }` (never returns the value). - `PUT` → upserts `{ value }` into the env block. - `DELETE` → removes the key. No-op when it wasn't set. Per-method scope: GET is treated as a write action so the single action name maps to one consistent scope; reading whether a secret is set is restricted to writers in any case.

Alerts Create

Alerts Destroy

Alerts List

Alerts Partial Update

Alerts Retrieve

Simulate a detector on an insight's historical data. Read-only — no AlertCheck records are created.

Alerts Update

Create, Read, Update and Delete annotations. [See docs](https://posthog.com/docs/data/annotations) for more information on annotations.

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Create, Read, Update and Delete annotations. [See docs](https://posthog.com/docs/data/annotations) for more information on annotations.

Create, Read, Update and Delete annotations. [See docs](https://posthog.com/docs/data/annotations) for more information on annotations.

Create, Read, Update and Delete annotations. [See docs](https://posthog.com/docs/data/annotations) for more information on annotations.

Create, Read, Update and Delete annotations. [See docs](https://posthog.com/docs/data/annotations) for more information on annotations.

Approval Policies Create

Approval Policies Destroy

Approval Policies List

Approval Policies Partial Update

Approval Policies Retrieve

Approval Policies Update

Cancel a batch export backfill.

Create a new backfill for a BatchExport.

ViewSet for BatchExportBackfill models. Allows creating and reading backfills, but not updating or deleting them.

ViewSet for BatchExportBackfill models. Allows creating and reading backfills, but not updating or deleting them.

Batch Exports Create

Batch Exports Destroy

Batch Exports List

Batch Exports Logs Retrieve

Batch Exports Partial Update

Pause a BatchExport.

Batch Exports Retrieve

Batch Exports Run Test Step Create

Batch Exports Run Test Step New Create

Cancel a batch export run.

Batch Exports Runs List

Batch Exports Runs Logs Retrieve

Batch Exports Runs Retrieve

Retry a batch export run. We use the same underlying mechanism as when backfilling a batch export, as retrying a run is the same as backfilling one run.

Batch Exports Test Retrieve

Unpause a BatchExport.

Batch Exports Update

Read-only access to parsed knowledge documents. Exposes hybrid search (``search``) and a drill-down window (``window``) so an agent (PHAI or MCP) can find and explore business knowledge chunks.

Read-only access to parsed knowledge documents. Exposes hybrid search (``search``) and a drill-down window (``window``) so an agent (PHAI or MCP) can find and explore business knowledge chunks.

Surfaces topics the support AI couldn't answer from the knowledge base. Two list shapes controlled by the ``ticket_id`` query param: - **per-ticket** (``?ticket_id=<uuid>``): individual gap rows for that ticket. - **aggregated** (no ``ticket_id``): gaps grouped by normalized topic with counts, for the Business knowledge suggestions panel.

Accept all pending suggestions for a normalized topic cluster.

Surfaces topics the support AI couldn't answer from the knowledge base. Two list shapes controlled by the ``ticket_id`` query param: - **per-ticket** (``?ticket_id=<uuid>``): individual gap rows for that ticket. - **aggregated** (no ``ticket_id``): gaps grouped by normalized topic with counts, for the Business knowledge suggestions panel.

Dismiss all pending suggestions for a normalized topic cluster.

Surfaces topics the support AI couldn't answer from the knowledge base. Two list shapes controlled by the ``ticket_id`` query param: - **per-ticket** (``?ticket_id=<uuid>``): individual gap rows for that ticket. - **aggregated** (no ``ticket_id``): gaps grouped by normalized topic with counts, for the Business knowledge suggestions panel.

Business Knowledge Sources Create

Business Knowledge Sources Destroy

Business Knowledge Sources List

Business Knowledge Sources Partial Update

Business Knowledge Sources Refresh Create

Business Knowledge Sources Retrieve

Business Knowledge Sources Text Retrieve

Approve a change request. If quorum is reached, automatically applies the change immediately.

Cancel a change request. Only the requester can cancel their own pending change request.

Change Requests List

Reject a change request.

Change Requests Retrieve

Manage CIMD verification tokens for an organization. A partner embeds the plaintext token in their CIMD metadata document as `verification_token` inside the `com.posthog` object (the legacy top-level `posthog_verification_token` field still works as a fallback). When PostHog fetches the metadata, matching the token links the partner app to this organization and grants a higher default rate limit for account provisioning. The plaintext value is only available on creation; we store a hash.

Manage CIMD verification tokens for an organization. A partner embeds the plaintext token in their CIMD metadata document as `verification_token` inside the `com.posthog` object (the legacy top-level `posthog_verification_token` field still works as a fallback). When PostHog fetches the metadata, matching the token links the partner app to this organization and grants a higher default rate limit for account provisioning. The plaintext value is only available on creation; we store a hash.

Manage CIMD verification tokens for an organization. A partner embeds the plaintext token in their CIMD metadata document as `verification_token` inside the `com.posthog` object (the legacy top-level `posthog_verification_token` field still works as a fallback). When PostHog fetches the metadata, matching the token links the partner app to this organization and grants a higher default rate limit for account provisioning. The plaintext value is only available on creation; we store a hash.

Manage CIMD verification tokens for an organization. A partner embeds the plaintext token in their CIMD metadata document as `verification_token` inside the `com.posthog` object (the legacy top-level `posthog_verification_token` field still works as a fallback). When PostHog fetches the metadata, matching the token links the partner app to this organization and grants a higher default rate limit for account provisioning. The plaintext value is only available on creation; we store a hash.

Check whether the authenticated user has access to PostHog Code.

Redeem a PostHog Code invite code to enable access.

Cohorts Activity Retrieve

Cohorts Add Persons To Static Cohort Partial Update

Cohorts All Activity Retrieve

Cohorts Calculation History Retrieve

Cohorts Create

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Cohorts List

Cohorts Partial Update

Cohorts Persons Retrieve

Cohorts Remove Person From Static Cohort Partial Update

Cohorts Retrieve

Cohorts Update

Cohorts Used In Retrieve

Mark a task-comment as complete. Sets completed_at and completed_by. 400 if the comment is not a task or is already complete.

Comments Count Retrieve

Comments Create

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Comments List

Comments Partial Update

Reopen a completed task-comment. Clears completed_at and completed_by. 400 if the comment is not a task or is already open.

Comments Retrieve

Comments Thread Retrieve

Comments Update

Appends a message to an existing conversation without triggering AI processing. This is used for client-side generated messages that need to be persisted (e.g., support ticket confirmation messages).

Cancel the conversation's in-progress LangGraph run.

Unified endpoint that handles both conversation creation and streaming. - If message is provided: Start new conversation processing - If no message: Stream from existing conversation

Delete a conversation.

Conversations List

Create-or-resume a sandbox conversation — the single sandbox session opener. With `content`, processes the turn (first message, in-progress follow-up, or terminal resume); without `content`, warms a sandbox that idles awaiting the first message. Returns the `(task, run)` handle the frontend opens SSE against. The conversation row is created on first use from the URL id.

Conversations Queue Clear Create

Conversations Queue Create

Conversations Queue Destroy

Conversations Queue Partial Update

Conversations Queue Retrieve

Conversations Retrieve

Record reviewer feedback on an AI reply, captured to the internal analytics project.

Update the status of multiple tickets in a single request. Only tickets belonging to the current team are affected; other-team UUIDs are silently ignored. Tickets already in the requested status are skipped.

Bulk update tags on multiple objects. PAT access: this action has no ``required_scopes=`` on the decorator — inheriting viewsets must add ``"bulk_update_tags"`` to their ``scope_object_write_actions`` list to accept personal API keys. Without that opt-in, ``APIScopePermission`` rejects PAT requests with "This action does not support personal API key access". Done per-viewset so granting ``<scope>:write`` for one resource doesn't leak access to sibling resources that share this mixin. Accepts: - {"ids": [...], "action": "add"|"remove"|"set", "tags": ["tag1", "tag2"]} Actions: - "add": Add tags to existing tags on each object - "remove": Remove specific tags from each object - "set": Replace all tags on each object with the provided list

Create a new outbound ticket and send the first message to the customer.

Conversations Tickets Create

Conversations Tickets Destroy

List tickets with person data attached.

Return the message thread for a ticket, ordered chronologically (paginated).

Conversations Tickets Partial Update

Post a reply or internal note to a ticket. With is_private=false, the reply is delivered to the customer via the ticket's channel (email, Slack, Teams, GitHub). With is_private=true, the message is stored as an internal note only visible to team members.

Get single ticket and mark as read by team.

Get total unread ticket count for the team. Returns the sum of unread_team_count for all non-resolved tickets. Cached in Redis for 30 seconds, invalidated on changes.

Handle ticket updates including assignee changes.

Conversations Views Create

Conversations Views Destroy

Conversations Views List

Conversations Views Retrieve

Create

Generate AI summary for a group of session recordings to find patterns and generate a notebook.

Generate AI individual summary for each session, without grouping.

Custom Property Definitions Create

Custom Property Definitions Destroy

Custom Property Definitions List

Custom Property Definitions Partial Update

Custom Property Definitions Retrieve

Custom Property Definitions Update

Custom Property Sources Create

Custom Property Sources Destroy

Custom Property Sources List

Custom Property Sources Partial Update

Custom Property Sources Retrieve

Custom Property Sources Update

Customer Journeys Create

Customer Journeys Destroy

Customer Journeys List

Customer Journeys Partial Update

Customer Journeys Retrieve

Customer Journeys Update

Customer Profile Configs Create

Customer Profile Configs Destroy

Customer Profile Configs List

Customer Profile Configs Partial Update

Customer Profile Configs Retrieve

Customer Profile Configs Update

Creates a new team-scoped template in the **target** project (URL) from a **team-scoped** source template in the same organization. Global and feature-flag templates return 400. Cross-organization or inaccessible sources return 404. Source and destination projects must differ (400 if equal). Conflicting `template_name` values on the destination are auto-suffixed with `(copy)`, `(copy 2)`, …

Dashboard Templates Create

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Dashboard Templates Json Schema Retrieve

Dashboard Templates List

Dashboard Templates Partial Update

Dashboard Templates Retrieve

Dashboard Templates Update

Bulk update tags on multiple objects. PAT access: this action has no ``required_scopes=`` on the decorator — inheriting viewsets must add ``"bulk_update_tags"`` to their ``scope_object_write_actions`` list to accept personal API keys. Without that opt-in, ``APIScopePermission`` rejects PAT requests with "This action does not support personal API key access". Done per-viewset so granting ``<scope>:write`` for one resource doesn't leak access to sibling resources that share this mixin. Accepts: - {"ids": [...], "action": "add"|"remove"|"set", "tags": ["tag1", "tag2"]} Actions: - "add": Add tags to existing tags on each object - "remove": Remove specific tags from each object - "set": Replace all tags on each object with the provided list

Dashboards Collaborators Create

Dashboards Collaborators Destroy

Dashboards Collaborators List

Copy an existing dashboard tile to another dashboard (insight, text card, or widget tile).

Dashboards Create

Dashboards Create From Template Json Create

Add a markdown text tile to a dashboard. Text tiles render as markdown blocks on the dashboard — useful as section headings, dividers, or annotations between insight tiles to give the dashboard structure.

Creates an unlisted dashboard from template by tag. Enforces uniqueness (one per tag per team). Returns 409 if unlisted dashboard with this tag already exists.

Soft-delete a single tile from a dashboard. Works for text, insight, and button tiles. The underlying Insight, Text, or ButtonTile object is preserved — only the dashboard tile is hidden. To delete the entire dashboard, use the dashboard delete endpoint instead.

Hard delete of this model is not allowed. Use a patch API call to set "deleted" to true

Dashboards List

Dashboards Move Tile Create

Dashboards Move Tile Partial Update

Dashboards Partial Update

Dashboards Reorder Tiles Create

Dashboards Retrieve

Run all insights on a dashboard and return their results.

Dashboards Run Widgets Retrieve

Dashboards Sharing List

Create a new password for the sharing configuration.

Delete a password from the sharing configuration.

Dashboards Sharing Refresh Create

Stream dashboard metadata and tiles via Server-Sent Events. Sends metadata first, then tiles as they are rendered.

Dashboards Update

Update the markdown body, layout, or color of an existing text tile on a dashboard.

Update the settings of existing widgets in place, atomically — config, name, and description. Each entry targets a widget by its tile_id and reuses the same write path as the dashboard PATCH endpoint. The widget_type is immutable. This edits widget settings only (config, name, description); tile placement (layouts, show_description) is a dashboard concern — use the dashboard PATCH endpoint or reorder_tiles for that. All updates succeed or fail together. To add new widgets, use the widgets/batch POST endpoint; to remove one, use delete_tile.

List registered dashboard widget types and per-type config_schema documentation for agents.

Add multiple widget tiles to a dashboard in one atomic request.

Data Color Themes Create

Data Color Themes Destroy

Data Color Themes List

Data Color Themes Partial Update

Data Color Themes Retrieve

Data Color Themes Update

List data modeling jobs which are "runs" for our saved queries.

Get the most recent non-running job for each saved query from the v2 backend.

List data modeling jobs which are "runs" for our saved queries.

Get all currently running jobs from the v2 backend.

Check if a database name is available.

Returns completed/non-running activities (jobs with status 'Completed'). Supports pagination and cutoff time filtering.

Returns failed/disabled data pipeline items for the Pipeline status side panel. Includes: materializations, syncs, sources, destinations, and transformations.

Returns the data ops overview dashboard ID for this team, creating it if it doesn't exist yet.