Cognee

Cognee

#467 most-used

Give your agents a knowledge graph built from your own data

ProductivityDocumentsDatabaseDeveloperAIAutomation

Cognee is an AI memory platform that turns your organisation's text into a queryable knowledge graph. Connect it to Actionist and your agents can add call transcripts, documents, and notes to named datasets, trigger the cognify process to build a structured knowledge graph, and run graph-completion or RAG searches to retrieve contextually accurate answers grounded in your own institutional memory — not a generic model.

Average time saved
14 hours
per person · per month
≈ 2 workdays back

Eliminates manual work. Agents eliminate the manual search through historical documents, call records, and policy libraries that knowledge workers perform before every decision — replacing fragmented search with a single structured memory query.

Schedule

What your Cognee agent runs on autopilot

A week of scheduled jobs your Actionist agent will execute on your behalf.

28Scheduled jobs
7Agents at work
24/7Always on
Agents
Wed–Fri
Wed
Thu
Fri
7a
8a
9a
10a
11a
12p
1p
2p
3p
4p
5p
6p
Multi-app workflows

Cognee × every other app you use

End-to-end automations that span multiple apps — each one a real business outcome.

6Workflows
6Apps spanned
~38 hrsSaved / week
6Personas served
For support
Featured4 apps

Knowledge-grounded ticket reply with approval gate

When a new support ticket arrives, the agent searches the support knowledge base and product documentation simultaneously in Cognee, combines the two most relevant results into a draft reply, and posts it for human approval in Slack before sending. Support staff review a pre-researched answer instead of starting from scratch.

~14 hrs

Time saved for your team — every week, on autopilot

The flow
Trigger·When a new support ticket lands in the shared Gmail inbox
Result
Draft suggested reply combining knowledge graph result and docs answerPost draft reply to #support-review for human approval before sending
The win
Saved per run
18 min
Runs / week
~45×
Every reply draws from your own resolved-case memory, not a generic model
Driven byCustomer Support Agent
ROI

Savings

What your team gets back — two angles: what you stop doing manually, and what that's worth.

Without Actionist

What you do manually today

With Actionist

What your agent runs for you

  • Sales
    25 min / week
    Manual CRM digging before each call

    Reps manually scroll through CRM notes, email history, and saved documents to piece together account context before a renewal call — often incomplete and always time-consuming.

    Sales Agent
    0 min
    Agent queries call memory graph before every renewal

    Before a renewal call, the sales agent searches years of call transcripts and playbooks in Cognee and delivers a pre-call brief to the rep's Slack 60 minutes before the call starts.

  • Marketing
    30 min / week
    Manual content deduplication and research

    Writers search through past published content to avoid repeating angles, then separately review customer feedback documents — a process that is often skipped under deadline pressure.

    Marketing Agent
    0 min
    Agent checks brand memory before every brief

    When a new content brief is created, the marketing agent queries Cognee for every past angle on that topic and every relevant customer signal, enriching the brief before the writer opens it.

  • Customer Support
    22 min / week
    Manual search through past ticket history

    Support agents search help desk history and product docs manually for each complex ticket — inconsistent retrieval quality and significant research time before a reply can be drafted.

    Customer Support Agent
    0 min
    Agent retrieves resolution from support knowledge graph

    When a ticket arrives, the support agent searches Cognee's knowledge graph for the closest matching past resolution and product documentation, posting a pre-researched draft for agent review within about a minute.

  • Human Resources
    40 min / week
    Manual policy document lookup per query

    HR staff search through shared drives and policy documents to answer each employee question — 10 minutes per query, multiplied across dozens of questions per week.

    Human Resources Agent
    0 min
    Agent answers policy questions from Cognee memory

    When an employee asks an HR question in Slack, the HR agent retrieves the relevant policy section from Cognee and replies in the thread with the policy name and clause — within about a minute.

  • Finance
    35 min / week
    Manual contract cross-reference during invoice review

    Finance staff manually compare invoice terms against vendor contracts and procurement policy documents during review cycles — a slow, error-prone process across high volumes of invoices.

    Finance Agent
    0 min
    Agent traces obligation chains through the compliance graph

    The finance agent searches the contracts and policy knowledge graph to identify obligation conflicts at invoice review time, surfacing chain-of-thought reasoning across multiple documents automatically.

  • Operations
    60 min / week
    Manual knowledge base maintenance

    Knowledge base managers manually upload documents, tag them, and keep search indexes current across multiple tools — a full-time curation burden that still leaves information stale.

    Operations Agent
    0 min
    Agent runs coordinated weekly knowledge graph refresh

    The operations agent batches all new content into Cognee datasets and triggers a single coordinated cognify run weekly, so every department's agents are querying current memory without manual curation.

  • Legal
    45 min / week
    Manual contract review against regulatory library

    Legal counsel manually compares new contract clauses against a regulatory reference library and past agreements — a time-intensive review that can delay contract execution by days.

    Legal Agent
    0 min
    Agent ingests contracts and checks compliance on arrival

    When a new contract is added, the legal agent ingests it into Cognee and searches the regulatory knowledge graph for conflicts immediately, flagging issues before the contract moves to signature.

+ 100s of other Cognee automations
Average time saved
26 hrs / person / month
Calculator

Calculate what your team saves

Team size
8 people
Hourly rate
$45 / hr
Hours saved / week
28
Hours saved / year
1,400
Annual ROI
$63,000

Based on Cognee's typical team usage — the visible tasks plus a few other automations the agent runs: ~3.5 hrs / person / week of admin work automated.

Connect

How to plug Cognee into Actionist

Pick the connection method that suits your environment.

Connect with your Cognee Cloud Base URL and API Key. Both are available on the API Keys page in your Cognee dashboard.

1
Open the Cognee API Keys page

Log in to your Cognee Cloud account and navigate to the API Keys page from the dashboard.

2
Copy your Base URL and API Key

Copy your Base URL (e.g. https://tenant-xxx.aws.cognee.ai) and your API Key. Both are displayed on the API Keys page.

3
Paste into Actionist and test

Paste both values into the Actionist connection form and click Test connection. Actionist will call the Cognee health endpoint to confirm the credentials are valid.

Credentials you'll need
Base URL*
Found on the API Keys page in your Cognee Cloud dashboard (e.g. https://tenant-xxx.aws.cognee.ai)
API Key*
Your Cognee API key — sent in the X-Api-Key header for all requests
Actions

12 actions your agent can call

Read and write operations available to your Actionist agent.

FAQs

Questions about Cognee + Actionist

How does Actionist connect to Cognee?
Go to the Apps tab, find Cognee, and click Connect. You will need two pieces of information from your Cognee dashboard: your Base URL (e.g. https://tenant-xxx.aws.cognee.ai) and your API Key. Both are found on the API Keys page inside your Cognee Cloud account. Paste them into the connection form and Actionist will run a health-check call to confirm the handshake before any actions run.
What is a Cognee dataset and how does the add-then-cognify workflow operate?
A dataset in Cognee is a named container for related text data. You first add text to a dataset with the Add Text to Dataset action, then trigger the Cognify action to build a knowledge graph from that dataset. Once cognified, the dataset's contents can be queried through the Search Memory action. You can maintain multiple independent datasets for different domains such as product documentation, customer interactions, or legal knowledge.
What are the three Cognee search types and when should I use each one?
Cognee Search supports three modes: GraphCompletion uses the knowledge graph to answer questions with graph-traversal reasoning; ChainOfThought adds step-by-step reasoning on top of the graph for complex multi-hop questions; RagCompletion uses classic retrieval-augmented generation for faster, context-window-based answers. For most agent queries, GraphCompletion gives the best balance of accuracy and speed. Use ChainOfThought when the query requires reasoning across multiple related concepts.
Can different Actionist agents use separate Cognee knowledge bases?
Yes. Because datasets are named containers, you can create one per knowledge domain — for example, a 'product-docs' dataset for support agents and a 'sales-playbook' dataset for sales agents. Each Actionist agent reads only the datasets relevant to its role. When you search, you pass the specific dataset names so the memory lookup stays scoped. This prevents knowledge bleed between departments.
How long does the Cognify step take and how should I schedule it?
The Cognify step that builds the knowledge graph can take several minutes depending on the volume of text in the dataset. Plan your agent schedules accordingly: add and cognify overnight or during off-peak hours, then run search queries during business hours once the graph is ready. The connection timeout is set to 10 minutes for the cognify call, so a dataset with many documents is still supported. Incremental adds to an existing dataset require a fresh cognify pass to update the graph.
What happens when I delete a Cognee dataset and can I recover the data?
Deleting a dataset removes it and all associated data permanently from Cognee. You can also delete individual data items within a dataset while keeping the dataset itself intact. Before deleting, confirm the dataset ID from your Cognee dashboard. Deletion is irreversible, so agents should always require a human approval gate before running a Delete Dataset action in production environments.
Can I keep Cognee memory up to date automatically on a schedule?
Yes. Because adding text and triggering cognify are separate steps, you can build a recurring pipeline that ingests new content on a schedule. For example: every Monday at 6 AM, add this week's new support tickets to the 'support-kb' dataset, then trigger cognify to update the knowledge graph. By Tuesday morning the updated memory is ready for support agents to query. This keeps your AI memory current without manual intervention.
How is Cognee different from a standard vector database or RAG pipeline?
Cognee is a knowledge graph memory layer, not a traditional vector database. It builds a graph of entities and relationships from your text, so search results reflect semantic connections rather than keyword proximity alone. This makes it better suited for questions that span multiple documents or require understanding relationships between concepts. For straightforward keyword lookups, RagCompletion mode behaves more like a vector search.
Get started

Connect your apps in minutes.

Start with a free instant demo, or talk to our team about a deployment designed for your business.