Using the Console

The console is where your organization's memory becomes visible. Everything your integrations ingest — every entity, fact, and relationship Honeycomb infers — can be inspected, questioned, and governed here, without writing a line of code. If the API is how memory gets in, the console is how you see what it became.

At a glance

• Ask is home — type a question, get a cited answer, and click every receipt back to its source.
• Insights surfaces what you didn't ask — contradictions, dropped commitments, and agent opportunities mined from the corpus.
• Org Intelligence models the org itself — find experts on any topic and see reporting lines inferred from how people actually work.
• Graph Explorer and Timeline show the memory as a living graph and as a chronology.
• Sources & Health is your ground truth — if a document appears there, it made it in.
• Admin & Access is where you mint API keys, manage Spaces and grants, and connect sources.

Signing in

Open your organization's console URL and sign in one of two ways:

1. Work email — enter your email, receive a 6-digit code, and type it in.
2. Google — one click with your Google workspace account.

After sign-in you land in your organization's workspace. Two things scope everything you see:

• Your organization — if you belong to more than one, the switcher in the sidebar changes the active org.
• The memory store — the isolated container holding all ingested knowledge. Every screen, and every API key you create, is scoped to the active store.

Note: The console always acts as you. Answers, digests, and document visibility respect your own access — you see what has been shared with you and what lives in org-wide Spaces, never someone else's private context. See Spaces and permissions.

Finding your way around

The sidebar has seven screens. Each answers a different question about your memory:

Note: A brand-new store shows a "memory is building" state on most screens until your first content lands and enrichment completes. Run the Quickstart to get something in.

Ask — answers with receipts

Ask is the home page and the product's core verb: a single search box over your organization's entire memory.

What you see. A question box with an effort dial (low / medium / high) and suggestion chips generated from your actual memory — real people, real projects, real open questions. Submit a question and the answer streams in with inline [n] citation markers.

What you can do:

• Read the answer, then check its work. Below the answer, a trace strip shows how it was assembled — what was searched, what was retrieved, how long it took. Real numbers, not hand-waving.
• Flip between Receipts and Graph. Receipts view lists the cited source documents as cards; Graph view draws the entities and relationships the answer touched.
• Open any receipt. Each card links to the underlying source with full provenance.
• Turn the dial. low is a fast single pass; high runs thorough multi-step retrieval for questions that span sources. Details in Ask.

Try it — verify a claim end to end:

1. Ask a question you know has a real source, e.g. What are the terms of the Meridian Health renewal?.
2. Read the answer and note the [1] citation marker.
3. In the Receipts view, open receipt [1] — you land on the exact document that grounds the claim.
4. If the answer says it can't find enough evidence, believe it: Honeycomb refuses rather than guesses.

Insights — what you didn't ask

Insights is a feed of things the mining engine found on its own, each grounded in evidence.

What you see. Cards filtered by type, with live counts:

What you can do:

• Expand the evidence. Every card shows how many memories ground it; expand to read them.
• Acknowledge or dismiss. Triage the feed — handled insights leave the queue.
• Mine on demand. The "Mine insights" button runs a fresh mining pass over the corpus instead of waiting for the next scheduled one.

Try it — triage a contradiction:

1. Filter the feed to Contradictions.
2. Open a card and expand its evidence — you'll see both conflicting statements with their sources.
3. Resolve it in the real world (or ingest a correcting document), then Acknowledge the card.

See Insights for how mining works and how to drive it from the API.

Org Intelligence — the org modelling itself

This screen shows what Honeycomb has inferred about your organization from how people actually communicate — not from an HR export.

What you see. Two widgets:

• Find the expert — type a topic, project, customer, or team and get a ranked list of people, each with a relative expertise score and the number of source documents backing it.
• Inferred org chart — reporting lines grouped by manager, derived from real interaction patterns.

What you can do. Click any person to open their profile page: what they own, who they report to and manage, their areas of expertise, and their recent activity — all with evidence.

Try it — prep for a meeting:

1. Type the customer's name into the expert search, e.g. Meridian Health.
2. Open the top-ranked person's profile.
3. Skim what they own and their recent activity — you now know who to loop in and what they last touched.

Note: Inferred structure improves with signal. A quiet corpus produces a sparse chart; a busy one converges on reality surprisingly fast.

Graph Explorer — memory, visually

The Graph Explorer renders your memory graph — people, teams, projects, companies, and the relationships between them — as an interactive canvas.

What you see. Nodes sized by importance, edges weighted by how often the relationship is reinforced. Edge opacity reflects confidence: relationships that stop being mentioned fade over time.

What you can do:

• Search for any person, team, or project to spotlight it.
• Filter by entity type to declutter.
• Toggle Neighbourhoods to shade the graph into its natural communities — the informal teams your org actually has.
• Include decayed edges to see relationships that have faded from disuse.
• Replay the graph forming, oldest to newest — watch your organization's knowledge assemble itself.
• Click any node to open its dossier: current facts, relationships, timeline, and recent mentions, with a link to the full entity page.

Try it — map a customer's blast radius:

1. Search for a customer, e.g. Meridian Health.
2. Click the node and skim the dossier — who's connected, through what.
3. Toggle Neighbourhoods to see which internal cluster owns the relationship.

Timeline — how knowledge unfolded

Timeline reconstructs a topic as a chronology: every relevant event in your memory, sorted by when it happened (not when it was ingested).

What you see. A search box; submit a topic and you get a vertical timeline of events, each with a source badge and the entities involved.

Try it — reconstruct a deal:

1. Search for the deal or project, e.g. the Meridian Health renewal.
2. Read the timeline top to bottom — first contact, the negotiation turns, the signature.
3. Click through any event's entities to their dossiers for the surrounding context.

This is temporal retrieval — the same capability the API exposes as mode: "temporal" on POST /api/v1/ask. See Search and recall.

Sources & Health — is the firehose alive?

The ground-truth screen for anyone building an integration. Two tabs.

Overview

• Corpus KPIs — entries, entities, relationships, memories, mentions.
• Ingest activity — a 7-day volume chart with a live/quiet badge, so a silent connector is visible at a glance.
• Enrichment backlog — the share of entries fully enriched, plus a live throughput strip (backlog, processing rate, ETA to drain) whenever a backlog exists.
• Per-source health — each source platform with its document count, last-seen time, and enrichment backlog. Also available raw via GET /api/v1/sources/health.
• Breakdowns — by source platform and by content type.

Documents

A filterable, paginated inspector over every ingested document. Filter by title search, source, content type, and enrichment status.

Click any row to open its provenance drawer — the full audit of what that one document became:

Try it — verify your integration is landing:

1. Ingest a test document via POST /api/v1/ingest (see the Quickstart).
2. Open Sources → Documents and search for its title.
3. Check its enrichment status — once complete, open the provenance drawer and confirm the entities and facts you expected.
4. If it's not there, check the Overview tab: a growing backlog means enrichment is catching up; a quiet firehose means your integration never sent it.

Admin & Access — keys, people, integrations

Admin is scoped to the active organization and memory store (the context header at the top tells you exactly which). Five tabs:

Task: mint an API key

1. Open Admin → API keys.
2. Name the key for its job (e.g. Production webhook) and create it.
3. Copy it immediately — it's shown exactly once and can't be retrieved again. Lost keys are revoked and re-minted, never recovered.
4. Use it as a bearer token: Authorization: Bearer hck_... on every API call.

Warning: A key can read and write the whole store it's scoped to. Keep it in a secrets manager, and revoke any key you can't account for — revocation is immediate.

Task: grant a teammate access to a Space

Access in Honeycomb is granted per Space, not per document — a grant lights up every document the Space contains, now and in the future.

1. Open Admin → Access.
2. Find the Space (Spaces are created automatically as content flows in; the org-wide Space always exists).
3. Add the teammate with a role, e.g. reader.
4. Their console — and any API calls acting on their behalf via the x-acting-user header — now include that Space's content.

Full model, roles, and the fail-closed routing rules: Spaces and permissions.

Next steps

• Quickstart — get your first document in and your first cited answer out.
• Ask — effort levels, scoped questions, and retrieval traces, from the API side.
• Connectors — keep the firehose fed continuously.
• Spaces and permissions — the access model behind everything the console shows you.
• API reference — every endpoint the console is built on top of.