Skip to main content
November 2026 release notes

Breaking changes

  • An object or object_list subfield nested inside another returns as an object or a list of objects, not JSON text. See Nested objects.
  • Search status, replay, export and exclusion lists answer only the caller that started the search, its team, or a team it was shared with. Other callers get 404. This covers GET /search/status/{task_id}, lookalike status, GET /job-status/{job_id} and replays.
  • An Idempotency-Key on POST /workflows/run names one run permanently, not for 24 hours. Reusing it returns that run with already_started: true. A run that could not be confirmed returns 503 with code run_admission_unconfirmed and its job_id. See Idempotency-Key header.
  • Clients regenerated from the API schema no longer include unused and app-only fields. Existing requests still work.

API & Data

The /v1 API
  • People and Company Intelligence accept field_confidence for per-field scores, justifications and sources. See Confidence Scoring.
  • Company Intelligence accepts full_org_chart to request the chart in its result. See Company Intelligence.
  • Every long-running /v1 request returns one Job that you poll, list, cancel and export the same way. See Async Jobs.
  • Workflows, Templates, Schedules, Monitors, Webhooks, Team skills, Connections, Feedback, Age Verification and balance have /v1 routes. See Route changes.
  • Schedule responses include the current owner_user_id; Team skills pages include policy actions for editing. See Schedules and Team skills.
  • A Workflow run is a Job. Start it with POST /v1/workflows/{workflow_id}/runs. See Run a Workflow.
  • Send Idempotency-Key when you start a Job. A retry returns the same Job and is not charged again. See Safe retries.
  • /v1 errors share one body with error_code, retryable and request_id. See Handling Errors.
  • POST /v1/files uploads CSV, JSON or NDJSON. Jobs take input_file_id, and exports return a file to download.
  • A Job’s charge_credits is null when the charge is not known yet or your role cannot see spend.
  • A /v1 file’s download_url starts with https://.
  • A single-record People or Company Intelligence Job returns the record in Job.result. A batch returns its output files there, without storage details.
  • Routes without /v1 work as before and are listed under Legacy.
  • Start Atlas investigations in batches and continue them with Idempotency-Key retries. Attached files must be uploaded by the caller: API keys can use only files uploaded by that same key, and people can use only their own uploads. See batch investigations and continuations.
  • Filter agent events by node and page backward with before. See Agent event stream.
  • Generate and poll report Jobs, page earlier reports, and read structured reports. See report generation, report job, history, and JSON.
  • API-key report revisions accept only evidence uploaded by that key; person callers can use their own uploads. Revisions require Idempotency-Key. See report revisions.
Other changes
  • Start an org chart with task: "org_chart" on POST /v1/atlas/investigations, and read saved structured output with GET /v1/atlas/investigations/{investigation_id}/output. See Org chart output.
  • POST /feedback reports wrong, outdated or missed data. See Send Feedback.
  • LinkedIn monitors take alert criteria per field and assess a change against the full stored profile. Editing criteria keeps the monitor’s history and baselines.
  • A monitor whose file upload stops partway reads status: failed after five minutes. A monitor whose API key is revoked, or whose creator leaves the team, pauses and sets disabled_reason. See Disabled reasons.
  • People Intelligence high runs take Team skills and connections.
  • People Intelligence Atlas runs take agent_trace and generate_report, both off by default. See Atlas integrations.
  • An Atlas run stops when the agents find the subject is under 18. It is not charged and returns nothing. See Declined runs.
  • GET /workflows/runs and the Atlas API add spend_access, which gives the reason when a cost is hidden from you.
  • GET /v1/atlas/investigations/{investigation_id}/identifiers.csv lists every email, phone, IP address, name, organization, account and domain an investigation found. See Atlas API.
  • Workflow blocks no longer need sequence_number or block_type. See Workflow definitions.
  • A Workflow run can finish while its result files are still saving. See Download results.
  • A Workflow run still starts when saving its input is slow. Input that cannot be kept in time returns 503 and is not charged.
  • When Sixtyfour’s data is briefly unreachable, every endpoint returns 503 with Retry-After. See Handling Errors.
  • Error bodies on the original routes carry a code and a correlation_id to quote to support. See Handling Errors.
  • exclude_public_ids on search is deprecated. Use exclude_entity_ids. See Search exclusions.
  • QA Agent is no longer documented. Existing integrations still work.

Platform

  • Team skills give research agents your team’s playbooks, and connections let them call your own APIs with a key they never see. Set both up under Agents in Settings. Available to organizations with each feature enabled.
  • On the Usage page, members see their own runs and spend. Team owners and organization admins see every member’s. An API key sees the costs its creator may see. A teammate’s cost shows as Hidden. See Roles & Permissions.
  • Share → Export → Identifiers CSV downloads every identifier an Atlas investigation found, one row each. See Export.
  • A Browser task in an Atlas run is charged only when it finds what it was asked for. See Billed Tools.
  • Transcription is a paid source. Atlas runs pay for it from their budget, and teams can turn it off or cap it per run.
  • An API key’s creator gets a warning before the key expires. See Expiration.
  • Organization admins and owners join a team before creating its API keys, because a key sees the team’s work. See Team scope.

Developer Tools

  • The MCP sixtyfour_send_feedback tool takes the same body as POST /feedback. See MCP setup.
October 2026 release notes

Breaking changes

  • Workflow, search and Atlas CSVs write lists and objects as JSON, and booleans as true or false. See CSV format.
  • Find Phone and reverse lookup webhooks carry their own task_type and counts. Update handlers that read emails_found from them.
  • API text fields have length limits. Longer values return 422. See Handling Errors.
  • GET /job-status/{job_id} and GET /workflows/runs/{run_id}/live_status share a polling limit of 300 requests a minute per organization.
  • OAuth token requests are rate-limited after repeated failed or concurrent sign-ins. See OAuth tokens.
  • Result download links are limited to 5 requests per second per organization, with a burst of 15.

API

Platform

  • Lattice (beta) turns rows of people and companies into one connected graph.
  • The find_email block can skip rows that already have an email, with skip_rows_with_existing_input.
  • Team owners and organization admins can choose and cap Atlas spending on billed tools, starting with the Browser.
  • Contact support to turn the high and xhigh tiers off for one team. See Roles & Permissions.

Developer Tools

September 2026 release notes

Breaking changes

  • GET /monitors/{monitor_id}, POST /monitors and GET /monitors/for-run/{workflow_run_id} no longer return rows. Read them with GET /monitors/{monitor_id}/rows.

API

Platform

  • Workflow rows that fail at a step skip later steps and are not charged for them. See Failed rows.
  • API keys and workflow schedules stay with their team after their creator leaves.
August 2026 release notes

API

  • Instagram research returns tagged posts with who tagged them and threaded comment replies, and searches users, hashtags and places.
  • LinkedIn research returns more company and job detail.

Platform

July 2026 release notes

API

  • The Age Verification API verifies a person’s age.
  • The Schedules API creates, updates and deletes Workflow schedules.
  • The xhigh tier is available for People Intelligence.
  • The high and xhigh tiers return structured findings from Instagram, TikTok, Reddit, X, YouTube and Facebook.
  • The micro tier returns up to 10 fields.
  • Large CSV exports no longer time out.

Platform

  • Age Verification is in the API Playground.
  • The LLM Output block has more models. See AI Model Pricing.
  • Workflows can export to Google Sheets. See Google Sheets integration.
  • The Workflow logs table shows who or what started each run.
  • Organizations that use only the API have a Developer Home page in the web app.
  • The API Playground has a Copy Agent Prompt button.

Developer Tools

June 2026 release notes

API

  • Bulk Intelligence takes a CSV or JSON file of up to 100,000 people or companies in one Job.
  • Check Balance returns your credit balance.
  • The micro tier is available for People and Company Intelligence and returns up to 5 fields.
  • Search exclusions keep people or companies out of /search/query and deep search, by inline IDs or reusable exclusion lists of up to 250,000 entries.
  • struct fields can be enum or list[enum], limited to the options you list. See Enums.
  • Confidence scores return the URLs each value came from.
  • Workflows read and write JSON, JSONL and NDJSON as well as CSV.
  • The high tier researches IP addresses.
  • Intelligence checks email deliverability.
  • OSINT research covers more email, phone and username sources.

Platform

  • The API Playground runs every endpoint in the browser, with code to copy and results inline.
  • Roles & Permissions control team access and account settings.
  • Agent chat runs a Workflow block by block and shows each block’s results inline.

Developer Tools

  • The MCP server adds People and Company Intelligence, Find Email, Find Phone and Check Balance.
  • MCP result polling is more reliable.
May 2026 release notes

API

  • Per-field confidence scores are available for People Intelligence results.
  • API keys can have an expiration date.
  • People Filter Search with several company filters, such as title, size and funding, matches people whose current company meets all of them.
  • Workflows that start with Reverse Email pass its domain and company fields to every later step.
  • The Verify Phone block checks numbers against more real-time data.

Platform

  • The Verify Phone block returns validity, carrier, line type, activity score, prepaid flag and country for a phone number.
  • Workflow templates add the categories Intelligence, Deanonymization, Finance, Trust & Safety, Background Checks, OSINT and AML/KYC. A template that needs a tier your organization lacks shows as Locked.
  • The web app has a Home page, and Settings has its own navigation section.
  • Signing up creates an organization for you.
  • Signup and login have captcha protection.
  • The Agent Prompt Library has example prompts for the Sixtyfour Agent.

Developer Tools

  • Agent Skills teach coding assistants such as Claude Code and Cursor to call the Sixtyfour API.
  • The docs have a full API Reference.
April 2026 release notes

API

Platform

  • Agent Preferences hold instructions and context the Sixtyfour Agent uses in every conversation.
  • YOLO Mode lets the Agent run tool actions without asking first.
  • The Agent reasons longer on complex requests.
  • Usage Logs export up to 10,000 rows of API history as CSV.
  • Free tier users can use the Agent, run searches and build Workflows.
  • You get an email when your credit balance runs low.
  • Workflow previews show each row as it finishes.
  • In the Workflow Editor, an Enrichment Struct can return nested fields.

Developer Tools

March 2026 release notes

API

Platform

  • SSO and SAML sign-in are available for enterprise customers.

Developer Tools