Stack Ai integration: operations for AI agents
Stack Ai connects to Type.com as an integration. This page lists the 91 operations documented for Claude and Codex in Type.com: 49 read, 30 write, 12 destructive.
About Stack Ai
No-Code Platform for Enterprise AI. Build, deploy, and scale AI workflows without code.
Categories: Developer Tools & IT, Productivity & Project Management
How to connect Stack Ai to Type.com
- Access
- Organization connection: one shared account for the workspace.
To add the connection, follow the steps in Space connections.
Documented operations
91 operations. Your connection's permissions determine which operations are available. API references describe the vendor surface; they do not expand your connection's credentials or access.
91 operations listed.
Endpoint get Run
- Read
Get Run
Get the status of a run, including the human-in-the-loop pause when waiting. How to use Use this to observe a legacy run or retrieve its pause context. Read state and pause from this response, not the v2 data.status/pendinginput fields. Terminal legacy success is COMPLETED.
Endpoint resume Run
- Write
Resume Run
Resume a paused run that is waiting on a human-in-the-loop response. How to use Read the current pause first and submit its token with the actual human answer. A RESUMED response can be nonterminal.
Endpoint run Deployed Flow
- Write
Run Deployed Flow
Run a flow and get the result. How to use Send the workflow input map as the entire JSON body. The flow must be published. Omit executionmode for inline JSON or SSE; request remote without SSE for a 202 background handle.
Endpoint list Knowledge Base Resources
- Read
List Resources
List files in the knowledge base with optional cursor pagination. How to use List resources and inspect processing status. Use cursor pagination and retain resourceid for detail reads. Do not treat a listed pending/parsed resource as indexed.
Endpoint upload Knowledge Base Resource
- Write
Create Resource
Upload a file directly to a knowledge base and start indexing. This endpoint allows API users to upload files directly without needing to use presigned URLs. The file is uploaded to storage and indexing is started automatically.
Endpoint get Knowledge Base Resource
- Read
Get Resource By Id
Get one file resource by ID from the knowledge base. How to use Poll this after upload/sync. indexed indicates readiness; error needs investigation. pendingdelete/deleted should not be treated as available search content.
Endpoint delete Knowledge Base Resource
- Destructive
Delete Resource By Id
Delete a file resource by ID. How to use Delete the specified resource from its KB. Verify both IDs before removing content used by downstream retrieval. A 204 has no body. Authentication and permissions Required permissions: kbs.delete.
Endpoint sync Knowledge Base
- Write
Synchronize Knowledge Base
Synchronize the knowledge base. How to use Requests a one-shot synchronization. A 202 means accepted, not that every resource is indexed. Inspect KB/resource status before querying updated source content. Authentication and permissions Required permissions: kbs.sync.
Endpoint list Knowledge Bases
- Read
Get Knowledge Bases
List all available knowledge bases for the specified flow. How to use Discover KBs available to the authenticated caller; select the returned knowledgebaseid. The v1 response is a bare array, so do not expect v2 cursor/envelope fields.
Endpoint create Knowledge Base
- Write
Create Knowledge Base
Create a new knowledge base. How to use Create a named content store, optionally selecting existing connection or website sources. Save knowledgebaseid, then upload or synchronize content. Catalog, source, embedding-model, and sharing policies can affect admission.
Endpoint get Knowledge Base
- Read
Get Knowledge Base By Id
Get a knowledge base by its ID in the current organization. How to use Read configuration and indexing totals before querying, changing, or deleting a KB. Existence does not mean all resources are indexed. Authentication and permissions Required permissions: kbs.read.
Endpoint update Knowledge Base
- Write
Update Knowledge Base
Update a knowledge base. How to use Send only fields to change and follow their nullability rules. Source and indexing configuration changes may need synchronization/reindexing; review resource readiness afterward. Authentication and permissions Required permissions: kbs.update.
Endpoint delete Knowledge Base
- Destructive
Delete Knowledge Base
Delete a knowledge base by its ID. How to use Removes the knowledge base. Check dependent workflows and retain necessary source material before deletion. A 204 response has no JSON body. Authentication and permissions Required permissions: kbs.delete.
Endpoint list Knowledge Bases V2
- Read
List Knowledge Bases
List knowledge bases available to the caller, newest first. How to use Discover KBs available to the authenticated caller; select the returned knowledgebaseid. Read items under data and follow nextcursor. Read successful resource payloads under data when a body is returned.
Endpoint create Knowledge Base V2
- Write
Create Knowledge Base
Create a new knowledge base. How to use Create a named content store, optionally selecting existing connection or website sources. Save knowledgebaseid, then upload or synchronize content. Catalog, source, embedding-model, and sharing policies can affect admission.
Endpoint get Knowledge Base V2
- Read
Get Knowledge Base
Get a knowledge base by its ID. How to use Read configuration and indexing totals before querying, changing, or deleting a KB. Existence does not mean all resources are indexed. Read successful resource payloads under data when a body is returned.
Endpoint update Knowledge Base V2
- Write
Update Knowledge Base
Update a knowledge base. How to use Send only fields to change and follow their nullability rules. Source and indexing configuration changes may need synchronization/reindexing; review resource readiness afterward.
Endpoint delete Knowledge Base V2
- Destructive
Delete Knowledge Base
Delete a knowledge base by its ID. How to use Removes the knowledge base. Check dependent workflows and retain necessary source material before deletion. A 204 response has no JSON body. Read successful resource payloads under data when a body is returned.
Endpoint get Knowledge Base Metadata Schemas V2
- Read
Get Knowledge Base Metadata Schemas
Get the user and dataloader metadata schemas for a knowledge base. How to use Inspect the user and dataloader metadata schema before interpreting file metadata. These metadata descriptions are not the content-search endpoint.
Endpoint list Knowledge Base Resources V2
- Read
List Resources
List all files in the knowledge base (flat). How to use List resources and inspect processing status. Use cursor pagination and retain resourceid for detail reads. Do not treat a listed pending/parsed resource as indexed.
Endpoint upload Knowledge Base Resource V2
- Write
Upload Resource
Upload a file directly and start indexing; poll the resource GET for status. How to use Upload multipart field file (singular). Indexing starts automatically. Save the resourceid from the 202 response and poll resource GET until indexed or error.
Endpoint get Knowledge Base Resource V2
- Read
Get Resource By Id
Get one file resource by ID. How to use Poll this after upload/sync. indexed indicates readiness; error needs investigation. pendingdelete/deleted should not be treated as available search content. Read successful resource payloads under data when a body is returned.
Endpoint delete Knowledge Base Resource V2
- Destructive
Delete Resource By Id
Delete a file resource by ID. How to use Delete the specified resource from its KB. Verify both IDs before removing content used by downstream retrieval. A 204 has no body. Read successful resource payloads under data when a body is returned.
Endpoint get Knowledge Base Resource Download Url V2
- Read
Get Resource Download Url
Get a signed URL to download the original file. How to use Returns a signed URL for the original file. Download separately without forwarding the StackAI Authorization header. Treat the URL as temporary sensitive access, not a permanent public link.
Endpoint query Knowledge Base V2
- Read
Query Knowledge Base
Search a knowledge base and return the most relevant results. How to use Search only after the resources you need are indexed. Choose semantic, keyword, or hybrid; topk is 1–20. This returns retrieval results, not a generated answer.
Endpoint sync Knowledge Base V2
- Write
Synchronize Knowledge Base
Start a one-shot synchronization of the knowledge base. How to use Requests a one-shot synchronization. A 202 means accepted, not that every resource is indexed. Inspect KB/resource status before querying updated source content.
Endpoint get Project Io Schema V2
- Read
Get Project Io Schema
Describe input keys and output nodes exposed by the selected project version. How to use Call before constructing run inputs or evaluation contracts. Match the selected saved version to the version you intend to execute.
Endpoint list Eval Datasets V2
- Read
List Eval Datasets
List matching project datasets; contract-drift warnings are available on detail reads. How to use Supply projectid. Use search or feeds compatibility filters to narrow the list.
Endpoint create Eval Dataset V2
- Write
Create Eval Dataset
Create a dataset with its input/output contract and advisory warnings. How to use Read project IO first and choose exactly one input key and output node ID for nodecontract. Create the dataset, save its ID, then append rows or import finished run traces.
Endpoint get Eval Dataset V2
- Read
Get Eval Dataset
Read one dataset and current contract-drift warnings. How to use Read before modifying a dataset or creating a regression experiment. Inspect contract-drift warnings and resolve differences from the workflow version you intend to test.
Endpoint update Eval Dataset V2
- Write
Update Eval Dataset
Update metadata or re-key rows to a replacement contract. How to use Use this for metadata or a deliberate contract replacement. Changing the contract can re-key stored rows; review the new mapping first. Omit unchanged fields.
Endpoint delete Eval Dataset V2
- Destructive
Delete Eval Dataset
Delete a dataset and its rows and experiments. How to use Deleting a dataset cascades to its rows and experiments. Retain required results before removal and verify the dataset belongs to the intended project. Authentication and permissions Required permissions: evals.delete.
Endpoint list Eval Dataset Rows V2
- Read
List Eval Dataset Rows
List dataset rows in position order. How to use Traverse in position order with offset/limit. Use the parent dataset ID together with returned row IDs for updates or deletes. Authentication and permissions Required permissions: evals.read.
Endpoint add Eval Dataset Rows V2
- Write
Add Eval Dataset Rows
Append rows as given; Location is the first created row. How to use Append 1–100 rows with input keys matching the dataset contract. Include expectedoutput for reference-based evaluation.
Endpoint import Eval Dataset Rows V2
- Write
Import Eval Dataset Rows
Import finished runs atomically, requiring evals.update and analytics.read. Inputs and expected outputs come from each run trace under the dataset contract. A run without a usable trace rejects the whole batch; Location is the first row.
Endpoint get Eval Dataset Row V2
- Read
Get Eval Dataset Row
Read one row after checking its dataset membership and project access. How to use Use to inspect the exact input and expected-output pair before changing it. The row must belong to the dataset in the path. Authentication and permissions Required permissions: evals.read.
Endpoint update Eval Dataset Row V2
- Write
Update Eval Dataset Row
Update row content, allowing an explicit null to clear expected output. How to use Omit expectedoutput to retain it; send explicit null to clear it. Keep input keys aligned to the parent dataset contract. Review the modified row before running experiments.
Endpoint delete Eval Dataset Row V2
- Destructive
Delete Eval Dataset Row
Delete one row belonging to the selected dataset. How to use Delete one row from its parent dataset. Verify parent membership and preserve examples needed for future regression coverage. Authentication and permissions Required permissions: evals.delete.
Endpoint list Evaluators V2
- Read
List Evaluators
List project evaluators with optional compatibility and exclusion filters. How to use Supply projectid. Filter by dataset/experiment compatibility, excluded IDs, or usesexpectedoutput. Monitors and direct run scoring require reference-free evaluators.
Endpoint create Evaluator V2
- Write
Create Evaluator
Create a validated evaluator and its first immutable configuration version. How to use Choose a built-in template or validate a judge configuration through preview first. Supply evaluatortype, config, and the project node contract.
Endpoint get Evaluator V2
- Read
Get Evaluator
Read an evaluator with its current version. How to use Read the evaluator and its current configuration version before using its ID in an experiment or monitor. Verify whether it expects reference output. Authentication and permissions Required permissions: evals.read.
Endpoint update Evaluator V2
- Write
Update Evaluator
Update metadata and version material changes to judge configuration. How to use Update metadata or judge configuration. Material config changes create a new immutable version; account for this when comparing historical scores.
Endpoint delete Evaluator V2
- Destructive
Delete Evaluator
Soft-delete an evaluator while retaining its historical scores. How to use Soft-delete an evaluator that should no longer be selected. Historical scores remain; deleting does not erase the prior evidence. Authentication and permissions Required permissions: evals.delete.
Endpoint preview Evaluator V2
- Read
Preview Evaluator
Judge up to three samples without saving the evaluator or score rows. How to use Test 1–3 samples without saving an evaluator or score rows. Use this to check rubric/config behavior before production use.
Endpoint list Evaluator Templates V2
- Read
List Evaluator Templates
Return built-in starting points for evaluator configuration. How to use Read supported starting configurations instead of guessing arbitrary config fields. Adapt a template to the project contract and preview it with representative samples.
Endpoint list Eval Experiments V2
- Read
List Eval Experiments
List a project's experiments and their aggregate scores. How to use Supply projectid and optionally datasetid/search. Use aggregates to locate an experiment, then inspect detail and row-level results before comparing quality.
Endpoint create Eval Experiment V2
- Write
Create Eval Experiment
Snapshot a dataset and validate the selected workflow version and evaluators. How to use Snapshot an existing dataset with a selected saved workflow version or draft and compatible evaluators. This creates the experiment; execution starts separately through jobs.
Endpoint get Eval Experiment V2
- Read
Get Eval Experiment
Get experiment metadata, progress, evaluator summaries and aggregate cards. How to use Read the experiment snapshot metadata, progress and aggregate cards. Aggregate success can coexist with individual row failures; inspect the rows as well.
Endpoint update Eval Experiment V2
- Write
Update Eval Experiment
Update metadata; omitted fields remain unchanged and null clears notes. How to use Update metadata only as permitted by the request schema. Omitted fields stay unchanged; null clears notes. This is not a replacement for creating a new experiment snapshot.
Endpoint delete Eval Experiment V2
- Destructive
Delete Eval Experiment
Delete an experiment and its row snapshots and scores. How to use Deletes the experiment with its row snapshots and scores. Export or record required results before removal. Authentication and permissions Required permissions: evals.delete.
Endpoint list Eval Experiment Rows V2
- Read
List Eval Experiment Rows
Page the experiment's rows with every row's score payloads nested together. How to use Page row snapshots with their nested score payloads. Inspect failures and score reasons; do not evaluate quality from aggregate metrics alone.
Endpoint get Eval Experiment Status V2
- Read
Get Eval Experiment Status
Read the latest job and current result counts and aggregates. How to use Use for latest-job overview and current counts/aggregates. For one accepted execution, poll its dedicated job URL so a later job does not change what you are observing.
Endpoint create Eval Experiment Job V2
- Write
Create Eval Experiment Job
Start a job on the experiment. With mode generate the job produces outputs for the experiment's rows and scores them in the same job when evaluator IDs are passed; with mode score it scores the existing outputs with the selected evaluators, at least one, and regenerates nothing.
Endpoint get Eval Experiment Job V2
- Read
Get Eval Experiment Job
A job is one accepted request to generate outputs for an experiment or to score its existing outputs. This route returns the job's lifecycle and, once it has finished, its outcome. Poll it until status is completed, completedwithfailures, failed or cancelled.
Endpoint get Eval Monitor Limits V2
- Read
Get Eval Monitor Limits
Get server-enforced sampling limits and monitor defaults. How to use Read enforced sampling limits and defaults before presenting a monitor-creation form. Do not permanently hard-code the policy in your client. Authentication and permissions Required permissions: evals.read.
Endpoint list Eval Monitors V2
- Read
List Eval Monitors
List project monitors with their seven-day score and alert rollups. How to use Supply projectid and inspect seven-day score/alert rollups. Read individual monitor configuration before changing sampling or notification behavior.
Endpoint create Eval Monitor V2
- Write
Create Eval Monitor
Create a monitor with one reference-free evaluator and bounded sampling. How to use Choose one reference-free evaluator and bounded sampling. Read monitor-limits first; save the monitor ID and inspect statistics after representative production runs.
Endpoint get Eval Monitor V2
- Read
Get Eval Monitor
Get one monitor's configuration and enabled state. How to use Read the current configuration and enabled state before a partial update. A monitor is a sampling policy, not an evaluation of every run. Authentication and permissions Required permissions: evals.read.
Endpoint update Eval Monitor V2
- Write
Update Eval Monitor
Update selected monitor fields, retaining omitted settings. How to use Send only settings to change and retain omitted values. Keep sampling within the current server limits; validate evaluator suitability when changing it.
Endpoint delete Eval Monitor V2
- Destructive
Delete Eval Monitor
Delete a monitor while retaining existing scores in run history. How to use Remove the monitor while retaining existing scores in run history. This is not an erasure of historical run evaluations. Authentication and permissions Required permissions: evals.delete.
Endpoint get Eval Monitor Score Stats V2
- Read
Get Eval Monitor Score Stats
Get windowed score trends and the monitor's current queue counters. How to use Choose 24h, 7d, 30d, or all. Inspect both trends and queue counters; low score volume may reflect sampling or queued work rather than lack of runs.
Endpoint get Eval Monitor Alert V2
- Read
Get Eval Monitor Alert
Read stored alert knobs, severity and resolved server defaults. How to use Read stored alert settings, current severity, and resolved defaults before tuning. Compare resolved values rather than assuming unset means disabled.
Endpoint update Eval Monitor Alert V2
- Write
Update Eval Monitor Alert
Tune alert settings; explicit null resets a knob to the server default. How to use Omit a knob to retain it; set it to null to restore the server default. Read back resolved settings after updating to verify the intended alert behavior.
Endpoint list Eval Monitor Sampled Runs V2
- Read
List Eval Monitor Sampled Runs
Page sampled runs and latest evaluator verdicts at the traversal's insertion cutoff. How to use Requires analytics.read as well as evals.read. Follow the cursor with unchanged filters; traversal preserves an insertion cutoff.
Endpoint list Eval Run Scores V2
- Read
List Eval Run Scores
Page a run's append-only score history, oldest first. How to use Read append-only score history oldest first. Requires analytics.read as well as evals.read. Follow cursors and compare evaluator versions before interpreting changed scores.
Endpoint create Eval Run Score V2
- Write
Create Eval Run Score
Append one score by evaluating an existing run with a reference-free evaluator. How to use Evaluate an existing run with a reference-free evaluator. Requires evals.run and analytics.read. Each call appends a score, so an ambiguous retry can add another history entry.
Endpoint get Eval Run Score V2
- Read
Get Eval Run Score
Retrieve a retained run score by its unique score ID. How to use Retrieve one retained score by score ID with evals.read and analytics.read. Use this to link an exact verdict in a review, rather than relying on the latest score.
Endpoint get Analytics Data Api
- Read
List Flow Runs
List flow run logs matching the given filters. How to use Use this for existing legacy exports. Read the response as a list. page starts at 0; dates should include a timezone. state uses uppercase legacy states and may contain comma-separated values.
Endpoint list Providers V2
- Read
List Providers
List integration providers available in the public catalog. How to use Use the returned provider IDs to discover actions and triggers; do not guess slugs. Advance offset/limit to inspect the full catalog. Listing a provider does not authorize use of any connection.
Endpoint get Provider V2
- Read
Get Provider
Get a single provider from the public catalog by id. How to use Read the descriptor for a provider selected from the catalog. Use its returned identifiers when fetching action and trigger collections. Authentication and permissions Required permissions: connections.read.
Endpoint list Provider Actions V2
- Read
List Provider Actions
List a provider's actions; fetch a single action for its input/output schema. How to use Choose an action ID here, then fetch its detail before invoking it. List entries are for discovery and need not contain the full input/output contract.
Endpoint get Provider Action V2
- Read
Get Provider Action
Get a provider action, including its input and output parameters. How to use Use the action-specific input and output parameter definitions to build inputs. Required fields, connection needs, and runtime behavior differ by action; there is no universal action payload.
Endpoint list Provider Triggers V2
- Read
List Provider Triggers
List a provider's triggers; fetch a single trigger for its parameter schema. How to use Discover the event sources an integration supports. Listing does not register, activate, or subscribe a workflow to a trigger. Authentication and permissions Required permissions: connections.
Endpoint get Provider Trigger V2
- Read
Get Provider Trigger
Get a provider trigger, including its input and output parameters. How to use Read the trigger-specific parameter contract for configuration in the supported workflow interface. This endpoint only reads a descriptor.
Endpoint run Provider Action V2
- Write
Run Provider Action
Run a provider action synchronously and return its outputs. How to use Select a compatible accessible connection and submit inputs matching this action. Omit connectionid only when supported by that action.
Endpoint list Connections V2
- Read
List Connections
List connections visible to the caller in their organization. Visibility follows RBAC: a non-admin sees connections they created, shared with them (directly or via a group), or carrying an org-level role; an admin sees the whole organization.
Endpoint list Webhook Subscriptions V2
- Read
List Webhook Subscriptions
List the organization's active webhook subscriptions, oldest first. How to use Inventory active subscriptions oldest first using offset/limit. The signing secret is not recoverable here. Subscription existence does not prove your runtime emits every accepted event type.
Endpoint create Webhook Subscription V2
- Write
Create Webhook Subscription
Register a webhook endpoint and return its signing secret. The secret is returned only here. It cannot be recovered later, so a caller that loses it has to delete the subscription and create a new one. How to use Configure and test an HTTPS receiver first.
Endpoint delete Webhook Subscription V2
- Destructive
Delete Webhook Subscription
Delete a webhook subscription: deliveries stop and it leaves the list. How to use Use the subscription ID returned by create/list. Deleting stops deliveries and removes it from active listing.
Endpoint list Runs V2
- Read
List Runs
List runs in accessible projects newest first. How to use Use this for run history or reconciliation. Follow nextcursor with unchanged filters. createdbyid is the exact returned actor ID; personal-key runs use the owner ID. Status filters use lowercase v2 values.
Endpoint start Run V2
- Write
Start Run
Start a run of a project version and return its current persisted state. How to use Send projectid and singular input. A 201 creates a run but does not mean it succeeded. Persist data.runid, then poll with analytics.read.
Endpoint get Run V2
- Read
Get Run
Get one of the organization's runs with its full payload. How to use Poll this after creation and inspect data.status. A 200 response can describe a failed run. The input/output may be null, and logging controls can block detail reads.
Endpoint cancel Run V2
- Destructive
Cancel Run
Request cancellation of an accessible run owned by this authenticated actor. Ownership matches the actor type and identity that started the run, including the distinction between a personal API key and a platform session.
Endpoint resume Run V2
- Write
Resume Run
Accept an asynchronous response to a paused ask-human step. How to use GET the paused run first, handle the human response, and send the issued resume token. Poll after 202. A 409 means state or pause/backend support is incompatible.
Endpoint list Skills V2
- Read
List Skills
List visible skills, with builtins first followed by organization skills. How to use Browse built-in skills first, followed by visible organization skills. Save the skill ID and fetch a pinned version when reproducibility matters.
Endpoint create Skill V2
- Write
Create Skill
Create and publish the first version from a multipart bundle or form fields. How to use Publish the first version using multipart form fields or bundle files. Use files (plural), repeat parts for multiple files, preserve relative paths, or upload a zip.
Endpoint get Skill V2
- Read
Get Skill
Retrieve the latest published skill and its public file manifest. How to use Read the current published instructions and public file manifest. Latest can change; use the version endpoint for an integration pinned to reviewed instructions.
Endpoint delete Skill V2
- Destructive
Delete Skill
Delete an owned skill and all versions; organization admins may delete any org skill. How to use This removes the skill and all versions. It is not a single-version unpublish. Ownership is required unless an organization administrator is deleting an organization skill.
Endpoint list Skill Versions V2
- Read
List Skill Versions
Page through a skill's version history, newest first. How to use List newest versions first with offset pagination. Choose an explicit version for an auditable release or rollback decision. Authentication and permissions Required permissions: skills.read.
Endpoint create Skill Version V2
- Write
Create Skill Version
Publish a new version, optionally carrying files from a previous version. How to use Publish a new revision. Repeat carrypaths for files to keep; baseversion selects the source version (current by default). Supplying a different name can rename the skill.
Endpoint get Skill Version V2
- Read
Get Skill Version
Retrieve one published version's instructions and public file manifest. How to use Fetch the exact 1-based version number. The response includes instructions and a public file manifest; retain this version alongside the consuming workflow configuration.
No operations match.
See How connections work for access and permissions, or browse all integrations.
Reference date: October 27, 2026. Stack Ai API docs.