REST API Reference for Serving Memory to AI Tools

Authentication

All API requests require a Bearer token in the Authorization header.

request headers
Authorization: Bearer your_token_here
Content-Type: application/json

All requests must include Content-Type: application/json. Responses use the same content type.

read_only

Read-Only Token

Access GET endpoints only. Perfect for AI tools that consume context.

full_access

Full Access Token

Access all endpoints — GET, POST, PATCH, DELETE. For management and automation.

Rate Limiting

API requests are limited to 100 requests per minute per token.

            429 response
            429 Too Many Requests
          
{
  "error": {
    "code": "RATE_LIMITED",
    "message": "Rate limit exceeded. Please try again later.",
    "status": 429
  }
}

Error Handling

All errors follow a consistent JSON structure with a code, message, and HTTP status.

error format
{
  "error": {
    "code": "ERROR_CODE",
    "message": "Human-readable description",
    "details": { ... },    // optional, for validation errors
    "status": 4xx
  }
}

Status	Code	Description
`401`	`UNAUTHORIZED`	Invalid or missing access token
`403`	`FORBIDDEN`	Insufficient permissions or inactive workspace
`404`	`NOT_FOUND`	Resource not found
`422`	`VALIDATION_ERROR`	Invalid request data (includes `details`)
`429`	`RATE_LIMITED`	Rate limit exceeded

Workspaces

List Workspaces

Returns all active workspaces for the current account.

GET/workspaces.json

Parameters

Name	Type	Description
page	integer	Page number (default: 1)

response200 OK
[
  {
    "id": 1,
    "name": "Project Alpha",
    "description": "Main project workspace",
    "memories_count": 42,
    "archived": false,
    "created_at": "2026-01-15T10:30:00Z",
    "updated_at": "2026-02-04T14:22:00Z",
    "url": "https://recuerd0.ai/workspaces/1"
  }
]

Pagination headers X-Page, X-Per-Page, X-Total, X-Total-Pages, and Link are included in all list responses.

Get Workspace

Returns a single workspace with memory count and pinned status.

GET/workspaces/:id.json

response200 OK
{
  "id": 1,
  "name": "Project Alpha",
  "description": "Main project workspace",
  "memories_count": 42,
  "archived": false,
  "created_at": "2026-01-15T10:30:00Z",
  "updated_at": "2026-02-04T14:22:00Z",
  "url": "https://recuerd0.ai/workspaces/1"
}

Get Workspace Context

Returns a compact "wake-up" snapshot of a workspace for AI agents — workspace metadata, a small set of orienting memories, and stats. When the requesting user has no pinned memories in the workspace, the endpoint returns the most recently updated memories instead of an empty list; context_source reports pins or recent. Supports HTTP caching via ETag.

GET/workspaces/:id/context.json

Query Parameters

Name	Type	Default	Description
limit	integer	10	Maximum memories to return (1–50).
include_body	boolean	true	Whether to include each memory's body content.
max_body_chars	integer	500	Maximum characters of body to return per memory (100–5000). Longer bodies are truncated.
category	string		Filter memories to one category. See Memory Categories.

response200 OK
{
  "workspace": {
    "id": 1,
    "name": "Project Alpha",
    "description": "Main project workspace",
    "memories_count": 42,
    "state": "active",
    "updated_at": "2026-04-01T12:00:00Z",
    "url": "https://recuerd0.ai/workspaces/1"
  },
  "memories": [
    {
      "id": 17,
      "title": "Architecture Notes",
      "source": "manual",
      "category": "decision",
      "tags": ["design", "core"],
      "pinned_at": "2026-03-28T09:14:00Z",
      "updated_at": "2026-04-01T11:42:00Z",
      "url": "https://recuerd0.ai/workspaces/1/memories/17",
      "body": "# Architecture\n\nThe system is split into…",
      "body_truncated": true,
      "links_count": 3
    }
  ],
  "context_source": "pins",
  "stats": {
    "total_memories": 42,
    "total_pinned": 3,
    "returned": 1,
    "returned_pinned": 1
  },
  "generated_at": "2026-04-06T10:00:00Z"
}

The pinned_memories array and stats.returned_pinned remain as deprecated aliases of memories and stats.returned.

Create Workspace

Creates a new workspace.

POST/workspaces.jsonfull_access

Parameters

Name	Type		Description
workspace[name]	string	required	Workspace name (max 100 characters)
workspace[description]	string		Workspace description

request
{
  "workspace": {
    "name": "New Project",
    "description": "A new workspace for the team"
  }
}

response201 Created
{
  "id": 2,
  "name": "New Project",
  "description": "A new workspace for the team",
  "memories_count": 0,
  "archived": false,
  "created_at": "2026-02-04T15:00:00Z",
  "updated_at": "2026-02-04T15:00:00Z",
  "url": "https://recuerd0.ai/workspaces/2"
}

Update Workspace

Updates an existing workspace.

PATCH/workspaces/:id.jsonfull_access

Parameters

Name	Type	Description
workspace[name]	string	Workspace name
workspace[description]	string	Workspace description

response200 OK
Returns the updated workspace object.

Archive Workspace

Archives a workspace. Archived workspaces become read-only.

POST/workspaces/:id/archive.jsonfull_access

response200 OK
Returns the workspace with "archived": true

Unarchive Workspace

Restores an archived workspace.

DELETE/workspaces/:id/archive.jsonfull_access

response200 OK
Returns the workspace with "archived": false

Memories

Memory Categories

Every memory carries exactly one category from a locked four-value enum. Category defaults to general and is used to filter list, search, and workspace context endpoints. Versions inherit the parent memory's category unless overridden.

Value	Label	Purpose
`decision`	Decision	Architecture choices, library picks, tradeoffs with stated reasoning
`discovery`	Discovery	Non-obvious findings — gotchas, root causes, patterns, library quirks
`preference`	Preference	User-stated rules ("always X", "never Y")
`general`	General	Catch-all and default

The category query parameter is supported on list, search, and workspace context endpoints to filter results to a single category.

Memory Links

Memory links — also called "tunnels" — are undirected, unlabeled "see also" connections between two memories within the same account. They cross workspace boundaries: a memory in your rails-app workspace can link to a memory in your mobile-app workspace. Use them to express that two memories cover the same topic without duplicating content.

Method	Path	Description
`GET`	`/workspaces/:workspace_id/memories/:memory_id/links.json`	List linked memories with their workspace info
`POST`	`/workspaces/:workspace_id/memories/:memory_id/links.json`	Create a link to another memory by id
`DELETE`	`/workspaces/:workspace_id/memories/:memory_id/links/:id.json`	Remove a link. `:id` is the other memory's id, not a join row id

Links are scoped to the same account — cross-account links are rejected with 422. Storage is undirected: linking A↔B is the same as B↔A and the server normalizes the pair so the link is only created once. There is no cardinality limit, links carry no labels or metadata, and deleting a memory cascades to remove all of its links. The DELETE endpoint identifies the link by the other memory's id, not a join row id. Memory and pinned-memory JSON responses include a links_count field showing how many links a memory has.

List Memory Links

Returns all memories linked to a given memory, with each linked memory's workspace embedded.

GET/workspaces/:workspace_id/memories/:memory_id/links.json

response200 OK
[
  {
    "id": 118,
    "title": "Auth strategy (mobile)",
    "category": "decision",
    "tags": ["auth", "mobile"],
    "source": "manual",
    "updated_at": "2026-03-22T11:05:00Z",
    "url": "https://recuerd0.ai/workspaces/2/memories/118",
    "workspace": {
      "id": 2,
      "name": "mobile-app",
      "url": "https://recuerd0.ai/workspaces/2"
    }
  },
  {
    "id": 204,
    "title": "Auth migration runbook",
    "category": "discovery",
    "tags": ["auth", "infra"],
    "source": "manual",
    "updated_at": "2026-04-02T08:30:00Z",
    "url": "https://recuerd0.ai/workspaces/5/memories/204",
    "workspace": {
      "id": 5,
      "name": "infra",
      "url": "https://recuerd0.ai/workspaces/5"
    }
  }
]

Create Memory Link

Creates an undirected link between this memory and another memory in the same account.

POST/workspaces/:workspace_id/memories/:memory_id/links.jsonfull_access

Parameters

Name	Type		Description
to_memory_id	integer	required	The id of the other memory to link to. Must belong to the same account.

request
{
  "to_memory_id": 118
}

response201 Created
{
  "id": 118,
  "title": "Auth strategy (mobile)",
  "category": "decision",
  "tags": ["auth", "mobile"],
  "source": "manual",
  "updated_at": "2026-03-22T11:05:00Z",
  "url": "https://recuerd0.ai/workspaces/2/memories/118",
  "workspace": {
    "id": 2,
    "name": "mobile-app",
    "url": "https://recuerd0.ai/workspaces/2"
  }
}

Returns 422 if to_memory_id belongs to a different account, references the same memory, or the link already exists.

Delete Memory Link

Removes a link between two memories. The path :id is the other memory's id — not a separate join row id.

DELETE/workspaces/:workspace_id/memories/:memory_id/links/:id.jsonfull_access

response204 No Content
Empty response body. Because links are undirected, removing the link from either side has the same effect.

List Memories

Returns all memories (latest versions only) for a workspace. Supports filtering by title pattern, tags, and source.

GET/workspaces/:workspace_id/memories.json

Parameters

Name	Type	Description
page	integer	Page number (default: 1)
per_page	integer	Items per page (1–100, default: 25)
title	string	Glob pattern for title matching. `*` matches any characters, `?` matches one
tags	string	Comma-separated tag list. Returns memories with ALL specified tags
source	string	Exact match on source field
category	string	Filter to one category. See Memory Categories.
sort	string	Sort field: `updated_at` (default), `created_at`, `title`
direction	string	Sort direction: `desc` (default), `asc`

examples
GET /workspaces/1/memories.json?title=Meeting*
GET /workspaces/1/memories.json?tags=api,design&sort=title&direction=asc
GET /workspaces/1/memories.json?category=decision&per_page=50

response200 OK
[
  {
    "id": 1,
    "title": "Meeting Notes",
    "version": 1,
    "source": "manual",
    "category": "decision",
    "tags": ["meetings", "q1"],
    "created_at": "2026-01-20T09:00:00Z",
    "updated_at": "2026-02-03T16:45:00Z",
    "url": "https://recuerd0.ai/workspaces/1/memories/1"
  }
]

Browse Memories

Returns memories across all active workspaces for the current account. Supports the same filtering parameters as List Memories. JSON-only endpoint.

GET/memories.json

Parameters

Name	Type	Description
page	integer	Page number (default: 1)
per_page	integer	Items per page (1–100, default: 25)
workspace_id	integer	Filter to a specific workspace
title	string	Glob pattern for title matching. `*` matches any characters, `?` matches one
tags	string	Comma-separated tag list. Returns memories with ALL specified tags
source	string	Exact match on source field
category	string	Filter to one category. See Memory Categories.
sort	string	Sort field: `updated_at` (default), `created_at`, `title`
direction	string	Sort direction: `desc` (default), `asc`

examples
GET /memories.json?title=*architecture*
GET /memories.json?tags=api,design&sort=title&direction=asc
GET /memories.json?workspace_id=1&source=manual

Only returns memories from active (non-archived, non-deleted) workspaces. Each memory includes its workspace in the response. Same response format and pagination headers as List Memories.

Get Memory

Returns a memory with its content. Supports line range parameters to read specific portions.

GET/workspaces/:workspace_id/memories/:id.json

Parameters

Name	Type	Description
line_start	integer	First line to return (1-based, inclusive). Default: 1
line_end	integer	Last line to return (1-based, inclusive). Default: last line

examples
GET /workspaces/1/memories/1.json
GET /workspaces/1/memories/1.json?line_start=1&line_end=20
GET /workspaces/1/memories/1.json?line_start=50&line_end=75

response200 OK
{
  "id": 1,
  "title": "Meeting Notes",
  "version": 1,
  "source": "manual",
  "category": "decision",
  "tags": ["meetings", "q1"],
  "created_at": "2026-01-20T09:00:00Z",
  "updated_at": "2026-02-03T16:45:00Z",
  "url": "https://recuerd0.ai/workspaces/1/memories/1",
  "content": {
    "body": "# Meeting Notes\n\nDiscussed Q1 goals...",
    "total_lines": 142,
    "line_start": 1,
    "line_end": 142
  },
  "links_count": 3,
  "workspace": {
    "id": 1,
    "name": "Project Alpha",
    "url": "https://recuerd0.ai/workspaces/1"
  }
}

The content object always includes total_lines, line_start, and line_end. When line range parameters are provided, body contains only the requested lines. Returns 422 if line_start > line_end.

Grep Mode Parameters

Name	Type	Description
mode	string	Set to `grep` to search within the memory
q	string	Search query required
context	integer	Lines before and after each match (0–10, default: 0)
before	integer	Lines before match, overrides `context` (0–10)
after	integer	Lines after match, overrides `context` (0–10)

examples
GET /workspaces/1/memories/1.json?mode=grep&q=architecture
GET /workspaces/1/memories/1.json?mode=grep&q=design&context=2

response200 OK
{
  "id": 1,
  "title": "Meeting Notes",
  ...
  "content": {
    "total_lines": 142,
    "matches": [
      {
        "line_number": 12,
        "line": "Discussed architecture decisions.",
        "context_before": ["## Design"],
        "context_after": ["The team agreed on microservices."]
      }
    ]
  }
}

In grep mode, the content object returns matches and total_lines instead of body. Returns 422 if q parameter is missing.

Workflow: grep, then fetch a window

For large memories, prefer a two-step read: call grep mode first to find each match's line_number, then call again with line_start/line_end to fetch a window of surrounding context.

step 1 — find matches
GET /workspaces/1/memories/1.json?mode=grep&q=TODO

response200 OK
{
  "content": {
    "total_lines": 412,
    "matches": [
      { "line_number": 47, "line": "TODO: revisit caching" }
    ]
  }
}

step 2 — fetch a 16-line window around line 47
GET /workspaces/1/memories/1.json?line_start=40&line_end=55

There is no head= or tail= parameter. Head is line_start=1&line_end=N; tail is line_start=(total_lines - N + 1)&line_end=total_lines. Because total_lines is echoed on every response, clients can compute tail without an extra round trip.

Create Memory

Creates a new memory with content.

POST/workspaces/:workspace_id/memories.jsonfull_access

Parameters

Name	Type	Description
memory[title]	string	Memory title (max 255 characters)
memory[content]	string	Memory body (Markdown)
memory[source]	string	Source identifier
memory[tags]	array	Array of tag strings
memory[category]	string	One of `decision`, `discovery`, `preference`, `general` (defaults to `general`). See Memory Categories.

request
{
  "memory": {
    "title": "API Documentation",
    "content": "# Overview\n\nThis document describes...",
    "category": "decision",
    "tags": ["docs", "api"]
  }
}

response201 Created
Returns the created memory object with content.

Update Memory

Updates an existing memory.

PATCH/workspaces/:workspace_id/memories/:id.jsonfull_access

Parameters

Name	Type	Description
memory[title]	string	Memory title
memory[content]	string	Memory body
memory[source]	string	Source identifier
memory[tags]	array	Array of tags
memory[category]	string	One of `decision`, `discovery`, `preference`, `general`. See Memory Categories.

response200 OK
Returns the updated memory object with content.

Delete Memory

Deletes a memory and all its versions.

DELETE/workspaces/:workspace_id/memories/:id.jsonfull_access

response204 No Content
Empty response body.

Memory Versions

Create Version

Creates a new version of a memory. Fields default to the parent version values if not provided.

POST/workspaces/:workspace_id/memories/:memory_id/versions.jsonfull_access

Parameters

Name	Type	Description
version[title]	string	Version title (defaults to parent)
version[content]	string	Version body (defaults to parent)
version[source]	string	Source identifier (defaults to parent)
version[tags]	array	Tags (defaults to parent)
version[category]	string	Category (defaults to the parent memory's category). See Memory Categories.

request
{
  "version": {
    "content": "# Updated Content\n\nRevised version..."
  }
}

response201 Created
{
  "id": 5,
  "title": "Meeting Notes",
  "version": 2,
  "source": "manual",
  "category": "decision",
  "tags": ["meetings", "q1"],
  "created_at": "2026-02-04T16:00:00Z",
  "updated_at": "2026-02-04T16:00:00Z",
  "url": "https://recuerd0.ai/workspaces/1/memories/5",
  "content": {
    "body": "# Updated Content\n\nRevised version..."
  },
  "workspace": {
    "id": 1,
    "name": "Project Alpha",
    "url": "https://recuerd0.ai/workspaces/1"
  }
}

Search

REST Search Memories

Full-text search across all memories in active workspaces. This REST endpoint accepts raw FTS5 query operators, requires at least three characters, and does not union tag matches.

GET/search.json?q=<query>

Parameters

Name	Type		Description
q	string	required	Search query (3–100 characters)
page	integer		Page number (default: 1)
workspace_id	integer		Filter results to a specific workspace
category	string		Filter results to one category. See Memory Categories.
mode	string		Response mode: `snippet` (default) or `grep`
context	integer		Lines before and after each match, like `grep -C` (0–10, default: 0). Only with `mode=grep`
before	integer		Lines before each match, like `grep -B` (0–10). Overrides `context`. Only with `mode=grep`
after	integer		Lines after each match, like `grep -A` (0–10). Overrides `context`. Only with `mode=grep`

Query Operators (FTS5)

Operator	Example	Description
Term	`architecture`	Matches documents containing the substring
AND	`architecture AND design`	Both terms must appear
OR	`meeting OR standup`	Either term can appear
NOT	`design NOT draft`	First term must appear, second must not
Phrase	`"project timeline"`	Exact phrase match
Column	`title:architecture`	Search only in title field
Column	`body:implementation`	Search only in body field
Grouping	`(meeting OR standup) AND notes`	Parentheses for precedence

release notes is a raw multi-term FTS5 query here. MCP and workspace search treat the same input as an exact phrase or exact tag. The paths differ by design: raw power for API clients, safe defaults for agents and the UI.

snippet mode response (default)200 OK
{
  "query": "architecture AND design",
  "total_results": 3,
  "results": [
    {
      "id": 1,
      "title": "Design Doc",
      "version": 1,
      "version_label": "v1",
      "has_versions": false,
      "category": "decision",
      "tags": ["design"],
      "source": "manual",
      "snippet": "Initial architecture overview...",
      "workspace": { ... }
    }
  ]
}

grep mode response200 OK
GET /search.json?q=architecture&mode=grep&context=1

{
  "query": "architecture",
  "total_results": 2,
  "results": [
    {
      "id": 1,
      "title": "Design Doc",
      ...
      "total_lines": 45,
      "matches": [
        {
          "line_number": 12,
          "line": "The architecture uses a layered approach.",
          "context_before": ["## System Design"],
          "context_after": ["Each layer communicates through interfaces."]
        }
      ],
      "workspace": { ... }
    }
  ]
}

Query must be 3–100 characters. Invalid FTS5 syntax returns a 422 with VALIDATION_ERROR.

In grep mode, each result includes matches and total_lines instead of snippet. FTS5 finds candidate memories, then line-level matching extracts specific lines with context.

API Reference

Authentication

Read-Only Token

Full Access Token

Rate Limiting

Error Handling

Workspaces

List Workspaces

Get Workspace

Get Workspace Context

Create Workspace

Update Workspace

Archive Workspace

Unarchive Workspace

Memories

Memory Categories

Memory Links

List Memory Links

Create Memory Link

Delete Memory Link

List Memories

Browse Memories

Get Memory

Create Memory

Update Memory

Delete Memory

Memory Versions

Create Version

Search

REST Search Memories

Product

Resources

Legal